Cover of Recommender Systems at Scale
DRM-free · Yours to keep forever
Data Science & ML

Recommender Systems at Scale

Collaborative filtering, embeddings and multi-stage ranking

By Houssam Kodad

PDF 268 pages Advanced English

One-time purchase

€29.95

VAT included
where applicable

Download sample
  • Instant download after purchase
  • Readable on any device
  • Free updates to this edition
  • Secure checkout

About this book

What's inside

Recommendations drive a huge share of engagement and revenue, yet the gap between a textbook matrix factorisation and a production recommender is enormous. This book covers the real architecture: candidate generation, ranking, and re-ranking, plus the embeddings and retrieval that make them fast. You'll learn to handle implicit feedback, the cold-start problem, and the offline/online evaluation that keeps a recommender honest.

What you'll learn

Skills you'll walk away with

  • Model implicit feedback rather than rare explicit ratings
  • Build collaborative filtering and matrix factorisation baselines
  • Learn user and item embeddings for retrieval
  • Design a two-stage candidate generation and ranking system
  • Use approximate nearest-neighbour search for fast retrieval
  • Tackle cold-start for new users and items
  • Evaluate offline and online without fooling yourself

Table of contents

9 chapters
  1. 01

    The Anatomy of a Recommender

    • · Retrieval, ranking, re-ranking
    • · Implicit vs explicit feedback
    • · Business objectives behind clicks
  2. 02

    Collaborative Filtering Foundations

    • · User and item neighbourhoods
    • · Matrix factorisation
    • · Handling sparsity
  3. 03

    Learning Embeddings

    • · From IDs to vectors
    • · Two-tower architectures
    • · Negative sampling that works
  4. 04

    Candidate Generation at Scale

    • · Approximate nearest neighbours
    • · Index building and refresh
    • · Blending multiple sources
  5. 05

    Ranking Models

    • · Features for ranking
    • · Gradient-boosted rankers
    • · Calibration of scores
  6. 06

    Re-Ranking and Business Rules

    • · Diversity and novelty
    • · Freshness and de-duplication
    • · Hard constraints and boosts
  7. 07

    The Cold-Start Problem

    • · New users and onboarding
    • · New items and content features
    • · Exploration strategies
  8. 08

    Evaluation You Can Trust

    • · Offline metrics and their lies
    • · Counterfactual evaluation
    • · Online A/B testing
  9. 09

    Serving and Feedback Loops

    • · Low-latency serving
    • · Logging for training data
    • · Avoiding feedback-loop bias

This is the full chapter list — exactly what you'll receive in the PDF.

More in Data Science & ML

Keep exploring this track

Cover of Feature Engineering for Machine Learning
Data Science & ML

Feature Engineering for Machine Learning

Crafting, selecting and serving features that move metrics

€27.95 View details
Cover of Practical MLOps: From Notebook to Production
Data Science & ML

Practical MLOps: From Notebook to Production

Packaging, deployment, monitoring and retraining that lasts

€32.95 View details
Cover of Time Series Forecasting in Practice
Data Science & ML

Time Series Forecasting in Practice

Classical models, gradient boosting and deep learning for demand and operations

€26.95 View details
Cover of Statistical Foundations for Data Scientists
Data Science & ML

Statistical Foundations for Data Scientists

Inference, experimentation and A/B testing done right

€24.95 View details
Cover of Gradient Boosting with XGBoost and LightGBM
Data Science & ML

Gradient Boosting with XGBoost and LightGBM

A practitioner's guide to winning with tabular data

€22.95 View details