Reinforcement Learning Insights
Aravind delves into the importance of model width and depth in self-supervised pre-training for images, contrasting supervised and reinforcement learning gains. He emphasizes the need for diverse environments like Imagenet for large models to impact reinforcement learning, highlighting the significance of scaling batch sizes in games like StarCraft and Dota.In this clip
From this podcast
Machine Learning Street Talk (MLST)
CURL: Contrastive Unsupervised Representations for Reinforcement Learning
Related Questions