Beyond exponential utility functions: A variance-adjusted approach for risk-averse reinforcement learning | IEEE Conference Publication | IEEE Xplore