Unlocking Personalization: A Deep Dive into Modern Recommendation Algorithms
summary
Walks through implemented pipelines on MovieLens 1M (≈1M ratings, ~6k users, ~3.8k movies) with runnable GitHub notebooks, showing heuristics (popularity corrected by a Bayesian average), content‑based TF‑IDF + cosine examples (Toy Story → Aladdin), collaborative filtering via matrix factorization, and a hybrid two-tower neural nets that produces user/item embeddings scored by dot product. Compares accuracy and tradeoffs, notes implicit vs explicit signals, and demonstrates end‑to‑end preprocessing, model training and deployment hooks — all code and solutions available.
AI-written summary · about
chapters (13)
- 0:00Introduction and agenda
- 2:22Motivation for RecSys
- 5:17Long tail theory
- 9:03Paradox of choice
- 12:27Brief history of RecSys
- 17:19Tutorial methodologies
- 18:52Input data types
- 21:28Heuristic-based recommendations
- 26:58Content-based filtering
- 30:16Collaborative filtering
- 35:32Hybrid recommenders
- 38:23Tutorial and dataset setup
- 52:24Practical implementation
related talks
37mA primer on recommendation systems.
32mLet's talk about Recommendation Systems!
46mBeyond Bag of Words A Practitioner’s Guide to Advanced NLP
01h 31mHands on - Build a Recommender system
23mML-Powered Search & Recommendation 101
57mHands-On with LLM-Powered Recommenders: Hybrid Architectures
58mRecommender Systems – A Walkthrough
01h 48mRecommendation engines explained