<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> This repository provides a comprehensive tutorial and practical guide, "Deep Recommendation Algorithm Practice (Wheat Book)," covering the evolution of recommendation systems from traditional cascading architectures to cutting-edge generative paradigms. It targets individuals with a machine learning background seeking to master recommendation algorithm principles and engineering practices, offering a structured learning path with real-world system implementation details.

How It Works

The content is structured into two main parts. The first part systematically covers traditional cascading recommendation techniques, including collaborative filtering, vector recall (I2I/U2I), sequence recall for user interest representation, feature crossing (second/high-order), multi-objective modeling, multi-scene modeling, and diversity-focused reranking methods. The second part delves into cutting-edge generative recommendation paradigms, exploring the foundations of LLMs, Scaling Laws for architecture exploration, end-to-end generative modeling frameworks like OneRec, and the application of diffusion models. This comprehensive approach provides a complete technical evolution narrative from established methods to future trends.

Quick Start & Requirements

The primary resource is the online reading address for the tutorial: https://datawhalechina.github.io/fun-rec/. Specific installation or execution commands are not detailed in the README. Prerequisites include a foundational understanding of machine learning.

Highlighted Details

• Provides a holistic view of recommendation system evolution, from traditional cascading architectures to advanced generative paradigms.
• Features in-depth exploration of cutting-edge topics such as Large Language Models (LLMs), Scaling Laws, Hardware-Aware architectures, Chain-of-Thought reasoning, and Diffusion Models applied to recommendation tasks.
• Includes a detailed conceptual walkthrough of building a production-grade recommendation system, covering system architecture, offline data processing pipelines, online serving flows, and deployment considerations.
• Offers practical insights into specific techniques like query completion generation, ad generation, and semantic unification for item indexing.

Maintenance & Community

The project is associated with the Datawhale learning community, offering a WeChat group and Knowledge Planet for discussion and content aggregation. Bilibili hosts related video content. Key contributors include Ruyi Luo and Bo Kang.

Licensing & Compatibility

Licensed under Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0). This license prohibits commercial use and requires derivative works to be shared under the same terms.

Limitations & Caveats

The README focuses on educational content rather than providing a runnable codebase. Specific implementation details for production systems are described conceptually, and direct code execution or setup instructions are absent. The CC BY-NC-SA 4.0 license restricts commercial application.