// the find
wepe/MachineLearning
Basic Machine Learning and Deep Learning
A Chinese-language study repo where one developer worked through classic ML algorithms from scratch — kNN, SVM, logistic regression, decision trees, Naive Bayes, PCA, GMM, and some early deep learning via Theano and Keras. It's a learning journal, not a library. The target audience is students who want to read clean, minimal implementations alongside the theory.
The from-scratch implementations are genuinely educational — the SVM has both an SMO and a QP solver, which is a useful contrast most tutorials skip. The decision tree covers both ID3 and C4.5, not just the easy one. Deep learning section includes CNN feature extraction and visualization, not just MNIST accuracy numbers. Everything is plain NumPy; no abstraction layers hiding what's actually happening.
The deep learning code depends on Theano, which was officially dead by 2020 — those notebooks will not run without significant patching. Last commit was June 2024 but the repo feels frozen circa 2015; there's nothing on transformers, attention, or anything post-2017. The README is entirely in Chinese with no English translation, which cuts off a large chunk of potential readers. C4.5 is explicitly marked as incomplete in the README, and CART was never added.