Online VNF Placement using Deep Reinforcement Learning and Reward Constrained Policy Optimization.
Ramy MohamedMarios AvgerisAris LeivadeasIoannis LambadarisPublished in: MeditCom (2024)
Keyphrases
- reinforcement learning
- optimal policy
- reward function
- partially observable environments
- policy search
- policy gradient
- function approximation
- reinforcement learning algorithms
- function approximators
- average reward
- state space
- concave convex procedure
- online learning
- control policy
- eligibility traces
- action selection
- markov decision process
- inverse reinforcement learning
- markov decision processes
- total reward
- real time
- markov decision problems
- reinforcement learning problems
- state action
- approximate dynamic programming
- actor critic
- state and action spaces
- partially observable
- partially observable domains
- action space
- partially observable markov decision processes
- model free
- control policies
- optimization algorithm
- policy evaluation
- agent receives
- optimal placement
- balancing exploration and exploitation
- exploration exploitation tradeoff
- learning agent
- temporal difference
- multi agent
- transition model
- agent learns
- dynamic programming
- saddle point
- rl algorithms
- reinforcement learning methods
- lagrange multipliers
- finite horizon
- infinite horizon
- decision problems
- sufficient conditions
- genetic algorithm