Spectral-Risk Safe Reinforcement Learning with Convergence Guarantees.
Dohyeong KimTaehyun ChoSeungyub HanHojun ChungKyungjae LeeSonghwai OhPublished in: CoRR (2024)
Keyphrases
- reinforcement learning
- convergence proof
- stochastic approximation
- actor critic
- convergence rate
- risk management
- high risk
- risk measures
- spectral analysis
- risk assessment
- function approximation
- robotic control
- temporal difference
- model free
- learning algorithm
- reinforcement learning algorithms
- state space
- convergence speed
- dynamic programming
- initial conditions
- machine learning
- multi agent
- multi agent reinforcement learning
- theoretical guarantees
- markov decision process
- multispectral images
- risk factors
- step size
- supervised learning
- optimal solution
- optimal policy
- decision making
- transfer learning