machine-learning
:earth_americas: machine learning tutorials (mainly in Python3)
What is machine-learning?
:earth_americas: machine learning tutorials (mainly in Python3)
Overview of machine-learning
What it is
machine-learning is a continuously updated repository that documents a personal journey on learning data science and machine learning related topics. The content is presented in Jupyter Notebook format and aims to strike a balance between mathematical notations, educational implementations from scratch using Python's scientific stack (numpy, numba, scipy, pandas, matplotlib, pyspark, etc.), and open-source library usage (scikit-learn, fasttext, huggingface, onnx, xgboost, lightgbm, pytorch, keras, tensorflow, gensim, h2o, ortools, ray tune, etc.). The repository is licensed under MIT and is open source.
Key features
- Covers a wide range of topics including deep learning, model deployment, operation research, reinforcement learning, time series, A/B testing, model selection, dimensionality reduction, recommendation systems, and more.
- Provides Jupyter Notebook tutorials with both from-scratch implementations and library-based examples.
- Includes end-to-end projects such as Kaggle competitions (e.g., Rossman Store Sales, Quora Insincere Questions Classification).
- Contains notes on advanced deep learning topics like Transformers, Contrastive Learning, LLM fine-tuning (PEFT LoRA, DPO, GRPO), and multi-modal models (CLIP, LLaVA).
- Offers practical guides on model deployment (FastAPI, Azure Kubernetes, AWS S3, Ray, VLLM) and quantization (FP8).
- Features A/B testing and causal inference tutorials (propensity score matching, difference-in-differences).
- Includes hyperparameter tuning with Ray Tune and Hyperband.
Use cases
This repository is intended for individuals learning data science and machine learning, from foundational concepts to advanced topics. It can be used as a reference for implementing algorithms from scratch, understanding mathematical underpinnings, and applying open-source libraries. The content is suitable for self-study, classroom supplementary material, or as a quick refresher on specific techniques such as recommendation systems, time series forecasting, or deep learning for tabular data.
Getting started
See the repository README for instructions on accessing the Jupyter Notebooks. The notebooks can be viewed via nbviewer links provided in the documentation listings.
License and pricing
The repository is licensed under MIT and is open source. No pricing information is stated.
Based on the project's public README and repository metadata.