Generative vs. Discriminative Models
Chris discusses the advantages of generative models over discriminative ones, highlighting their superior sample complexity even with limited data. He references upcoming research that reaffirms earlier theoretical findings about generative models' performance. Additionally, the conversation touches on the limitations of sequential models in capturing the hierarchical nature of language, suggesting that while they can manage nested structures, they may still fall short in fully representing language complexity.In this clip
From this podcast
NLP Highlights
04 - Recurrent Neural Network Grammars, with Chris Dyer
Related Questions