PAC-Bayes control: learning policies that provably generalize to novel environments

Mathematics

• Learning Control 100%
• Bayes 61%
• Policy 56%
• Control Policy 32%
• Generalise 31%
• Neural Networks 23%
• Generalization 16%
• Upper bound 16%
• Obstacle Avoidance 14%
• Minimise 13%
• Grasping 13%
• Supervised Learning 11%
• Stochastic Gradient 10%
• Leverage 10%
• Gradient Descent 10%
• Robot 10%
• Hardware 10%
• Robotics 10%
• Learning Algorithm 9%
• Relative Entropy 9%
• Stochastic Control 9%
• Machine Learning 9%
• Nonlinear Dynamics 8%
• Convex Optimization 8%
• Programming 7%
• Analogy 7%
• Learning 7%
• State Space 6%
• Costs 6%
• Optimization Problem 5%

Engineering & Materials Science

• Neural networks 35%
• Convex optimization 20%
• Supervised learning 20%
• Collision avoidance 19%
• Entropy 16%
• Learning algorithms 16%
• Robotics 14%
• Machine learning 14%
• Robots 12%
• Computer hardware 11%
• Costs 7%