How to Build a Winning Deep Learning Recommender System | Grandmaster Series E5

In this recommendation system challenge, the goal was to use a dataset based on millions of real anonymized accommodation reservations to come up with a strategy for making the best recommendation for their next destination, all in real-time. Watch the video to learn how more!

Subscribe to our Youtube channel to see a new Grandmaster Series episode each month.

If you have any questions during the video, you can submit them through chat. We will try to provide answers throughout and at the end of the episode.

Video Chapters:
00:00 – Intro to Episode Five
09:50 – Matrix Factorization Ensemble Overview
15:00 – Model One, MLP with Session-Based Matrix Factorization
16:18 – Model Two, GRU with MultiStage Session-based Matrix Factorization
20:00 – Model Three, XLNet with Session-based Matrix Factorization
24:34 – Combing All Three Models into One Ensemble
30:00 – Data Augmentation Approach
33:06 – NVIDIA Merlin is an End-2-End Library for GPU-Accelerated Recommender Systems
35:53 – Q&A Session

Six Additional Resources:

About our presenters:

Host Jim Scott, head of developer relations, data science, at NVIDIA. Over his career, he has focused on enabling business to solve the most complex engineering and data related problems. His expertise in blending business needs with technology to drive innovation has influenced every major industry.

Chris Deotte, senior data scientist at NVIDIA. Chris has a Ph.D. in computational science and mathematics with a thesis on optimizing parallel processing. Chris is a 4x Kaggle grandmaster.

Jean-Francois Puget (CPMP on Kaggle), Distinguished engineer at NIDIA. JFP holds a PhD in machine learning and has published over 70 scientific papers in peer reviewed conferences and journals. He is 2x Kaggle grandmaster.

Gabriel Moreira is a Senior Researcher at NVIDIA Merlin team working on Deep Learning for Recommender Systems, which was the focus of his Phd. He has previously worked for many years as Lead Data Scientist and Software Engineer.

Jiwei Liu is a senior data scientist at NVIDIA working on NVIDIA RAPIDS data science framework. Jiwei received a PhD degree in electrical and computer engineering. Jiwei is a 2x Kaggle grandmaster specialized in tabular data modeling.

[rtdcseng]

Os brasileiros estão mandando bem na NVIDIA. Show de bola

[chrisperidis]

Thank you very much for this nice and informative presentation. I would like to ask a general question about neural netowork training. Is it a better approach to increase the number of neurons per layer or to to increase the number of layers and thus the depth of the network, for achieving better accuracy results?

[ezazakhtar]

Waiting for the next video! This was so cool! Thank you!