Stochastic Optimization for Spectral Risk Measures

12/10/2022
by   Ronak Mehta, et al.
0

Spectral risk objectives - also called L-risks - allow for learning systems to interpolate between optimizing average-case performance (as in empirical risk minimization) and worst-case performance on a task. We develop stochastic algorithms to optimize these quantities by characterizing their subdifferential and addressing challenges such as biasedness of subgradient estimates and non-smoothness of the objective. We show theoretically and experimentally that out-of-the-box approaches such as stochastic subgradient and dual averaging are hindered by bias and that our approach outperforms them.

READ FULL TEXT

Please sign up or login with your details

Forgot password? Click here to reset