TensorFlow 1.2 study notes from a Chinese ML student

What it does

This repo is a Chinese-language companion to Google’s Udacity deep-learning course (UD730), rebuilt for TensorFlow 1.2. It walks through four lessons — from logistic regression up through CNNs and RNNs — with both theory notes and practice markdown files. The author also links to side repos for NumPy, matplotlib, sklearn, and a “watermelon book” (周志华’s ML classic) concept list.

The interesting bit

The candor. The README opens with “frameworks are tools” and ends with a 400×400 QR code begging donations for a quad-1080Ti rig. That tension — serious pedagogy meets broke grad student — gives the project more personality than most MOOC clones.

Key highlights

• Covers four lesson arcs: logistic classification → deep nets → convnets → sequence models
• Each lesson pairs a theory note with a “practical” (实践) exercise file
• Includes a TensorFlow install “pitfall log” (踩坑日志) — useful for the era
• Links to video and subtitle downloads for the Udacity course
• CC-BY-NC-ND licensed

Caveats

• Frozen in time: “refactored for TensorFlow 1.2” means pre-eager, pre-Keras-integration TF
• Appendix notes (NumPy, matplotlib, sklearn) marked “待完善” — unfinished
• No code visible in the README; actual notebooks/scripts live in subdirectories

Verdict

Worth a look if you’re a Mandarin speaker trying to make sense of legacy TensorFlow tutorials, or curious how self-directed ML education looked in China circa 2017. Skip if you need modern TF/PyTorch or English-first explanations.