Becoming a deep learning researcher without a PhD, graph neural network(GNN), time series, recommender system with Kyle Kranen - The Data Scientist Show#028
The Data Scientist Show - Daliana Liu
00:00
Navigating Challenges in Graph Neural Networks and Fraud Detection Models
The chapter explores the complexities of training graph neural networks (GNNs) on massive graphs like social networks and discusses techniques such as partitioning and sampling to handle the scale of data effectively. It highlights the importance of combating overfitting and data set imbalance in graph models, particularly in fraud detection scenarios, emphasizing the need for careful sampling strategies. Additionally, it delves into utilizing graph networks for interpretability, visualizing relationships, and addressing challenges like data drift in dynamic social networks for fraud detection.
Transcript
Play full episode
