It Just Takes Two: Scaling Amortized Inference to Large Sets
The episode discusses Antoine Wehenkel et al.'s paper, "It Just Takes Two: Scaling Amortized Inference to Large Sets." The hosts explain how Neural Posterior Estimation (NPE) uses pair training on sets of size two to learn representations that can be applied to any larger set size. This method decouples representation learning from posterior modeling, offering scalability and improved performance across various tasks.