// the find
wangshusen/DeepLearning
A university course (CS583) on deep learning, distributed as PDF slides and LaTeX lecture notes covering everything from logistic regression through Transformers, GANs, RL, and NAS. It's teaching material, not a library or tool — you're getting someone's grad-level course, not runnable code.
The breadth is real: it goes from ML fundamentals all the way to Vision Transformers, federated learning, and neural architecture search in a single coherent sequence. The lecture notes (LaTeX source included for BP, DRL, parallel computing) are more thorough than the slides alone and show the math properly. Coverage of adversarial robustness — both evasion and poisoning attacks — is rare in free course material at this level. Having slides and notes linkable by topic makes it easy to drop into a specific subject without reading the whole thing.
Last updated May 2021, which is a long time in this field — no coverage of diffusion models, LLM fine-tuning, RLHF, or anything post-BERT that actually matters in 2024. The repo is almost entirely PDFs with almost no executable code; the one Jupyter notebook (logistic regression) is the exception that proves the rule. It's Keras-focused at a time when PyTorch has become the de facto research standard, so the code examples that do exist will feel dated to anyone working in modern ML. No exercises, no problem sets, no way to test understanding beyond reading slides.