Estimating Uncertainty with Implicit Quantile Network

Inspired by Implicit Quantile Network from the reinforcement learning literature, this work aims to repurpose it for uncertainty estimation in supervised learning settings by modeling the entire distribution of the error.