PinSage - Graph Convolutional Neural Networks for Web-Scale Recommender Systems | Paper Explained
❤️ Become The AI Epiphany Patreon ❤️ ► https://www.patreon.com/theaiepiphany
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
In this video, I do a deep dive into the PinSage paper!
It was the first application of the GNN as a huge scale recommender system such as the one at Pinterest.
You'll learn about:
✔️All the nitty-gritty details behind PinSage
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
✅ paper: https://arxiv.org/pdf/1806.01973.pdf
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
⌚️ Timetable:
00:00 Intro to PinSage and Pinterest
02:40 High-level overview of engineering innovation
05:40 MapReduce framework - a quick overview
08:00 Quick overview of the algorithm
09:10 High-level overview of algorithmic innovation
10:15 Tasks and related work
13:00 Problem setting (bipartite graph)
14:40 Detailed explanation of the algorithm
18:00 Neighborhood sampling and importance pooling
21:45 Model training and supervised data
23:50 Max margin ranking loss function explanation
29:18 Negative (hard) samples and curriculum learning
35:30 Pin features and baselines description
39:00 Offline evaluations and metrics (hit-rate, MRR)
43:00 Analysis of the embedding space (cosine sim)
45:15 User studies and A/B testing
49:00 Embedding space visualizations (t-SNE)
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
💰 BECOME A PATREON OF THE AI EPIPHANY ❤️
If these videos, GitHub projects, and blogs help you,
consider helping me out by supporting me on Patreon!
The AI Epiphany ► https://www.patreon.com/theaiepiphany
One-time donation:
https://www.paypal.com/paypalme/theaiepiphany
Much love! ❤️
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
💡 The AI Epiphany is a channel dedicated to simplifying the field of AI using creative visualizations and in general, a stronger focus on geometrical and visual intuition, rather than the algebraic and numerical "intuition".
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
👋 CONNECT WITH ME ON SOCIAL
LinkedIn ► https://www.linkedin.com/in/aleksagordic/
Twitter ► https://twitter.com/gordic_aleksa
Instagram ► https://www.instagram.com/aiepiphany/
Facebook ► https://www.facebook.com/aiepiphany/
👨👩👧👦 JOIN OUR DISCORD COMMUNITY:
Discord ► https://discord.gg/peBrCpheKE
📢 SUBSCRIBE TO MY MONTHLY AI NEWSLETTER:
Substack ► https://aiepiphany.substack.com/
💻 FOLLOW ME ON GITHUB FOR COOL PROJECTS:
GitHub ► https://github.com/gordicaleksa
📚 FOLLOW ME ON MEDIUM:
Medium ► https://gordicaleksa.medium.com/
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
#pinsage #graphneuralnets #recommendersystems
Видео PinSage - Graph Convolutional Neural Networks for Web-Scale Recommender Systems | Paper Explained канала Aleksa Gordić - The AI Epiphany
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
In this video, I do a deep dive into the PinSage paper!
It was the first application of the GNN as a huge scale recommender system such as the one at Pinterest.
You'll learn about:
✔️All the nitty-gritty details behind PinSage
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
✅ paper: https://arxiv.org/pdf/1806.01973.pdf
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
⌚️ Timetable:
00:00 Intro to PinSage and Pinterest
02:40 High-level overview of engineering innovation
05:40 MapReduce framework - a quick overview
08:00 Quick overview of the algorithm
09:10 High-level overview of algorithmic innovation
10:15 Tasks and related work
13:00 Problem setting (bipartite graph)
14:40 Detailed explanation of the algorithm
18:00 Neighborhood sampling and importance pooling
21:45 Model training and supervised data
23:50 Max margin ranking loss function explanation
29:18 Negative (hard) samples and curriculum learning
35:30 Pin features and baselines description
39:00 Offline evaluations and metrics (hit-rate, MRR)
43:00 Analysis of the embedding space (cosine sim)
45:15 User studies and A/B testing
49:00 Embedding space visualizations (t-SNE)
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
💰 BECOME A PATREON OF THE AI EPIPHANY ❤️
If these videos, GitHub projects, and blogs help you,
consider helping me out by supporting me on Patreon!
The AI Epiphany ► https://www.patreon.com/theaiepiphany
One-time donation:
https://www.paypal.com/paypalme/theaiepiphany
Much love! ❤️
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
💡 The AI Epiphany is a channel dedicated to simplifying the field of AI using creative visualizations and in general, a stronger focus on geometrical and visual intuition, rather than the algebraic and numerical "intuition".
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
👋 CONNECT WITH ME ON SOCIAL
LinkedIn ► https://www.linkedin.com/in/aleksagordic/
Twitter ► https://twitter.com/gordic_aleksa
Instagram ► https://www.instagram.com/aiepiphany/
Facebook ► https://www.facebook.com/aiepiphany/
👨👩👧👦 JOIN OUR DISCORD COMMUNITY:
Discord ► https://discord.gg/peBrCpheKE
📢 SUBSCRIBE TO MY MONTHLY AI NEWSLETTER:
Substack ► https://aiepiphany.substack.com/
💻 FOLLOW ME ON GITHUB FOR COOL PROJECTS:
GitHub ► https://github.com/gordicaleksa
📚 FOLLOW ME ON MEDIUM:
Medium ► https://gordicaleksa.medium.com/
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
#pinsage #graphneuralnets #recommendersystems
Видео PinSage - Graph Convolutional Neural Networks for Web-Scale Recommender Systems | Paper Explained канала Aleksa Gordić - The AI Epiphany
Показать
Комментарии отсутствуют
Информация о видео
6 января 2021 г. 0:10:56
00:50:51
Другие видео канала
Feed-forward method | Neural Style Transfer #510k subscribers | joining Google DeepMind, updates, AMADay 25: Open NLLB - filtering HBS (Pt 2)Fastformer: Additive Attention Can Be All You Need | Paper ExplainedDay 24: Open NLLB - back from China, analyzing spikes, preparing HBS run (Pt 2)Day 6: Meta NLLB - data filtering (Pt. 4)Day 24: Open NLLB - back from China, filtering HBS data (Pt 3)Day 14: Open NLLB - Eval of our first run (English, Turkish, Hindi) (Pt 2.)The Vesuvius challenge breakthrough with Luke FarritorDay 4: Training 600M NLLB - data preps (Pt. 2)DeepMind's Android RL Environment - AndroidEnvHamel Husain - Building LLM Apps in ProductionDay 28: Open NLLB - debugging fuzzy dedup, training fasttext LID (Pt 3)Day 10: Open NLLB - evaluation data, filtering (Pt 3.)Day 18: Open NLLB - data loading document, GitHub tasks (Pt 1 cont.)DeepMind's TacticAI: an AI assistant for football tactics | Petar VeličkovićJarvis for Images! (demo) - run locally, no external APIsDay 15: Open NLLB - ALTI+, detecting hallucinations (Pt 2)Day 29: Open NLLB - training fasttext LID (Pt 1)LLaMA 2 w/ Thomas Scialom (LLaMA 2 lead)Day 19: Open NLLB - compute grant, downloading HBS parallel corpora (Pt 1)