In this work, we address risk-averse Bayes-adaptive reinforcementlearning. We pose the problem of optimising the conditional value at risk (CVaR) of the total return in Bayes-adaptive …
I Greenberg, Y Chow… - Advances in Neural …, 2022 - proceedings.neurips.cc
… of any risk-averse policy gradient (PG) method. We demonstrate improved riskaversion in … , including in scenarios where standard risk-averse PG completely fails. Our results and …
… Reward Design and Risk Modeling To train riskaverse agents, we propose an asymmetric reward function design such that good behavior receives small positive rewards and risky …
Y Shen, MJ Tobia, T Sommer… - Neural …, 2014 - ieeexplore.ieee.org
… We conclude (see Table 1) that most of the subjects are riskaverse for positive and risk seeking for negative TD errors. The result is consistent with previous studies from the economics …
… risk-aversereinforcementlearning can be used to hedge options. We apply a state-of-the-art risk-averse … of agents each characterized by a different riskaversion, so to be able to span …
… Although this choice strategy was not selected for, we rigorously prove that riskaversion … optimal reinforcementlearning can directly explain complex behaviors such as riskaversion …
… as riskaversion increases. This paper is organized as follows: in Section 2 we introduce the ReinforcementLearning … of the risk measures adopted and Multi-Objective Reinforcement …
O Mihatsch, R Neuneier - Machine learning, 2002 - Springer
… We provide a new theory of risk-sensitive control, 2. formulate reinforcementlearning algorithms within this framework which require only minor changes of already known and widely …
… In summary, we have presented a novel model of evolved reinforcementlearning agents, … optimal reinforcementlearning can directly explain complex behaviors such as riskaversion…