Abdullah Al Mamun

The Evaluation of RecSys — Part 3: The Deep Learning Era (NCF, Wide & Deep, DeepFM, DIN, DLRM, AdaTT)

Wed, 12 Mar 2025 00:00:00 +0000

Part 3 of the RecSys series. Traces the deep-learning revolution in RecSys from 2016 to 2023 — Neural Collaborative Filtering, Wide & Deep, DeepFM, Deep Interest Network, DLRM, and AdaTT. Architectures, intuition, where each one wins.

The Evaluation of RecSys — Part 2: Factorization Machines and XGBoost

Tue, 11 Mar 2025 00:00:00 +0000

Part 2 of the RecSys series. Factorization Machines generalize matrix factorization to arbitrary feature spaces, and XGBoost brings non-linear ranking via gradient-boosted trees. We cover the math, loss functions, strengths, and the limitations that drove the field toward deep learning.

The Evaluation of RecSys — Part 1: From Content-Based Filtering to Matrix Factorization

Sat, 01 Mar 2025 00:00:00 +0000

Part 1 of a deep-dive series on the evolution of recommendation systems. Covers content-based filtering, collaborative filtering (user/item), and matrix factorization — with loss functions, intuition, and where each technique breaks down.