Risk-sensitive and minimax control of discrete-time, finite-state Markov decision processes

By S. Coraluppi, S. Marcus. 1999

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Safe Model-based Reinforcement Learning with Stability Guarantees (2017, background): “In risk-sensitive RL, one specifies risk-aversion in the reward [7].”
AI Safety Gridworlds (2017, background): “A promising approach for guaranteeing baseline performance is risk-sensitive RL (Coraluppi and Marcus, 1999), which could be combined with distributional RL (Bellemare et al.”
Nonparametric Representation of Policies and Value Functions: A Trajectory-Based Approach (2002, background): “This is closely related to robust nonlinear controller design techniques based on the idea of control [11, 12] and risk sensitive control [13, 14].”
Robust Reinforcement Learning (2005, unknown): “ally derived by using dynamic programming (Imafuku, 1999; Coraluppi & Marcus, 1999 ).”
A Survey of Preference-Based Reinforcement Learning Methods (2017, background): “Furthermore, PbRL can currently not be used to perform risk-averse optimization (Coraluppi and Marcus, 1999; Geibel, 2001).”
A comprehensive survey on safe reinforcement learning (2015, background): “In many works, risk is related to the stochasticity of the environment and with the fact that, in those environments, even an optimal policy (with respect the return) may perform poorly in some cases (Coraluppi and Marcus, 1999; Heger, 1994b).”
Safe Exploration of State and Action Spaces in Reinforcement Learning (2012, background): “An example of such an approach is worst-case control where the worst possible outcome of R is to be optimized (Coraluppi & Marcus, 1999; Heger, 1994).”
Safe Exploration in Finite Markov Decision Processes with Gaussian Processes (2016, background): “In risk-sensitive reinforcement learning, the goal is to maximize the expected return for the worst case scenario [5].”
Risk-Sensitive Reinforcement Learning Applied to Control under Constraints (2005, background): “Most risk-sensitive RL approaches are concerned with the variance of the return, or with its worst outcomes, (e.g. Coraluppi & Marcus, 1999; Heger, 1994; Neuneier & Mihatsch, 1999), see also the discussion in section 3.”
Safe Exploration Algorithms for Reinforcement Learning Controllers (2018, methodology): “Several techniques exist to implement both minimax and risk-sensitive methods [25]; however, there are limitations as far as exploration is considered.”