Why Reinforcement Learning Is So Hard to Scale

In episode 995 of the SuperDataScience Podcast, Jon Krohn discusses with Jazmia Henry, Member of Technical Staff for AI and Machine Learning at Collide. Jazmia explains the concept of 'burstiness' in reinforcement learning models. She details how RL models can cause GPU usage to spike intermittently due to calculations being done on the CPU, leading to challenges in compute allocation and potential system failures if not managed correctly.

Видео Why Reinforcement Learning Is So Hard to Scale канала Super Data Science: ML & AI Podcast with Jon Krohn