DSP1 min read
IIR Filters and Quantization Error
A study note on how finite precision changes the behavior of recursive digital filters.
Working notes
Notes
Shorter records for ideas still being worked through: concepts, equations, implementation details, and experiments.
Experiment Logs1 min read
Experiment Log: Conditioning a Video Diffusion Model
An early record of how affordance conditioning changes temporal consistency in generated driving scenes.
Learning record
Learning timeline
- PostDeep Learning
Transformer 架构与训练详解
从整体架构、注意力机制到优化器与大规模预训练，系统梳理 Transformer 的组成和训练过程。
- PostReinforcement Learning
Bellman Optimality as a Planning Lens
A compact note on why the Bellman equation is more than a recurrence: it is a way to organize decisions under uncertainty.
- NoteDSP
IIR Filters and Quantization Error
A study note on how finite precision changes the behavior of recursive digital filters.
- PostWorld Model
What I Mean by a World Model
A working definition for connecting perception, temporal prediction, and action in embodied learning systems.
- NoteExperiment Logs
Experiment Log: Conditioning a Video Diffusion Model
An early record of how affordance conditioning changes temporal consistency in generated driving scenes.
- PaperWorld Model
Learning Latent Dynamics for Long-Horizon Video Prediction
- PaperAutonomous Driving
Planning-oriented Autonomous Driving