Deep Learning for Recommender Systems
summary
Survey of practical deep-learning recipes for recommendation: learns dense item vectors using Prod2Vec / item embeddings (skip‑gram on purchase sequences), uses autoencoder collaborative filtering to reconstruct user histories, and extracts image/audio features with pretrained nets like VGG / GoogLeNet or raw-CNNs. Builds a session model with GRUs, trains with ranking losses, and reports about 30% improvement over traditional session baselines; includes a hierarchical GRU to carry state across sessions for personalization.
AI-written summary · about
chapters (10)
- 0:00Intro
- 0:33Recommender Systems
- 2:38Why Deep Learning?
- 5:12Complex Architectures
- 6:33Research directions in DL-RecSys
- 12:06Geometry of the Embedding Space
- 12:432c for Recommendations
- 17:40Autoencoders for recommendation
- 22:53RNN-based machine learning
- 25:59Personalized Session-based RecSys
related talks
42mDeep Learning Recommender Architecture: Master Thesis to ...
29mFurther enhancements of recommender systems using deep learning
37mDeep Recommender Systems
21mNeural Networks for Recommender Systems
32mLet's talk about Recommendation Systems!
18mBuilding an Incrementally Trained, Local Taste Aware, Global Deep Learned Recommender System Model
21mUnleashing the Power of Data Science
30mLessons from building deep learning recommendation systems