A four-thousand-star filing cabinet for ad-tech research

What it does This repository is a curated archive of PDF papers and presentations focused on computational advertising. Organized into folders like Embedding, Budget Control, Deep Learning CTR Prediction, and Factorization Machines, it serves as a categorized reading list for ad-tech practitioners. The author, 王喆, updates it dynamically based on papers encountered in industry work.

The interesting bit Despite containing almost no code, the repo has attracted roughly 4,400 stars, likely because it bundles hard-to-find industry papers—like Airbnb’s KDD-best-paper embedding system and Alibaba’s DIEN model—with foundational classics such as the original MapReduce and Bigtable papers.

Key highlights

• Covers the full ad-tech stack: optimization methods (FTRL, Hogwild), topic models (LDA), tree models, and embedding techniques (Word2Vec, Node2Vec, DeepWalk).
• Includes industry-specific operational topics often missing from generic ML lists, such as budget pacing and real-time bid optimization from LinkedIn and DSP perspectives.
• Houses foundational distributed-systems papers—Google’s “Three Papers” on MapReduce, GFS, and Bigtable—alongside modern deep-learning CTR models.
• Links to related practitioner resources, including a Spark-based CTR estimation repo and separate recommender-system paper lists.

Caveats

• This is a bibliography, not a codebase: the repository hosts PDFs and links, not runnable tools or libraries.
• All materials are sourced from the open internet and hosted directly, with a note to contact the author if copyright issues arise.
• Scope is driven by the curator’s own work history, so the mix of foundational theory and industry papers reflects personal practice rather than a systematic academic survey.

Verdict Worth bookmarking if you are an ad-tech engineer or researcher building recommender, bidding, or pacing systems; skip it if you are looking for executable code or a rigorously peer-reviewed survey.