– Stage 2: Reranker model - Feature Selection & Engineering
8
– Multilingual Transfer Learning
9
– Task 3: Next Product Title Generation Solution
10
– Q&A Session
11
– Q&A: Challenges of Using Embedding
12
– Q&A: Unexpected Findings
13
– Q&A: Critical Model Improvements
14
– Q&A: Differences from OTTO Multi-Objective RecSys Competition
Description:
Dive into this 53-minute episode of the Grandmaster Series featuring the NVIDIA Merlin team's winning approach at the Amazon KDD Cup 2023 competition. Explore their innovative two-stage pipeline solution for building a high-performing e-commerce product recommendation system. Learn about cutting-edge techniques including transfer learning, embedding models, and multilingual approaches. Gain insights into candidate generation, co-visitation matrices, product embedding training, feature selection and engineering, and next product title generation. Benefit from a Q&A session covering embedding challenges, unexpected findings, critical model improvements, and comparisons with other competitions. Access additional resources on accelerating Pandas DataFrames, session-based recommender systems, and NVIDIA's data science solutions.
Mastering Multilingual Recommender Systems - Grandmaster Series