Aman's AI Journal • Recommendation Systems

Introduction

overview; retrieval; ranking

Embeddings

neural colab filtering; matrix factorization

Candidate Generation

notation; content-based filtering; code deep-dive; collaborative filtering; code deep-dive; retrieval

Ranking / Scoring

overview; scoring with the candidate generator?; objective function for scoring; position/selection bias

Re-ranking

overview; freshness; diversity; fairness

Calibration

platt scaling; isotonic regression; Bayesian calibration

Popular Architectures

Wide and Deep; Deep and Cross; DCN V2; NCF; DeepFM; AutoInt; DLRM; DHEN; GDCN; Two Towers

Search

lexical search; semantic search; hybrid search

Building a Music Recommendation System using PySpark

code deep-dive; data pre-processing; data aggregation; data split; spin up your model!

Evaluation Metrics and Loss Functions for Recommender Systems

evaluating candidate generation; evaluating candidate ranking; A/B testing; loss functions

Content Moderation for Recommender Systems

videos, text, and image moderation; early and late moderation

GNNs for Recommender Systems

GraphSage; Edge GNNs; use-cases

Transformers for Recommender Systems

PinnerFormer; ItemSage

Challenges of Building a Recommender System

result reproducibility; offline-online mismatch; oscillating outputs; slow convergence

Summary of Relevant Papers

popular recommendation systems in production demystified

Biases in Recommender Systems

clickbait bias; duration bias; position bias; popularity bias; measuring and mitigating positional bias; single-interest bias

Cold Start Problem

types of cold start; mitigation strategies

Multi-Armed Bandit

exploration; exploitation; contextual bandit

Multi-objective Optimization

LinkedIn use-case; overview

- Google's course on Recommendation Systems - Coursera: Music Recommender System Project - Coursera: DeepLearning.AI's specialization - Recommender system from learned embeddings - Google's Recommendation Systems Developer Crash Course - ALS introduction by Sophie Wats - Matrix Factorization - Recommendation System for E-commerce using Alternating Least Squares (ALS) on Apache Spark - Personalized Re-ranking for Recommendation - Dot Product Wiki

If you found our work useful, please cite it as:

{
  author        = {Chadha, Aman and Jain, Vinija},
  title         = {Recommender Systems Primer},
  howpublished  = {\url{https://www.aman.ai}},
  year          = {2022},
  note          = {Accessed: 2022-07-01},
  url           = {www.vinija.ai}
}

A. Chadha, V. Jain, Notes for Recommendation Systems https://www.aman.ai, 2022, Accessed: July 1 2022.