Optimizing a Neural Network Architecture for a Reinforcement Learning Agent
summary
The talk discusses optimizing the neural network architecture of an agent that learns via reinforcement learning. It covers techniques for modifying network structure and parameters to improve the agent’s learning performance in reinforcement-learning tasks.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 1:50Neural architecture search
- 5:23Architecture search approaches
- 7:31NAS with controller
- 11:07DARTS and supernets
- 13:38Reinforcement learning basics
- 20:25RL in Atari
- 23:18Search spaces and tuning
- 26:09Hypernetworks for NAS
- 29:45Experimental results
- 33:22Conclusion
- 34:07Q&A
related talks
31mDesigning game AI with reinforcement learning
34mOffline Reinforcement Learning and Model-Based Optimization
09mLightning Talk: Intellectual Cognitive Development in Artificial Intelligence
01h 17mTraining intelligent game agents using deep reinforcement learning
24mDesigning a Video Game AI with Reinforcement Learning
18mMaximizing Luck in Reinforcement Learning
02mDesigning a Video Game AI
24mIntelligent system optimizations