Autoregressive Quantile Flows for Predictive Uncertainty Estimation
This work addresses the need for flexible and efficient uncertainty estimation in machine learning applications, offering incremental improvements to existing flow-based methods.
The authors tackled the problem of representing probability distributions over high-dimensional data by proposing autoregressive quantile flows, a class of normalizing flow models trained with a novel objective that avoids expensive Jacobian calculations and supports new neural architectures, resulting in improved probabilistic predictions on tasks like time series forecasting and object detection.
Numerous applications of machine learning involve representing probability distributions over high-dimensional data. We propose autoregressive quantile flows, a flexible class of normalizing flow models trained using a novel objective based on proper scoring rules. Our objective does not require calculating computationally expensive determinants of Jacobians during training and supports new types of neural architectures, such as neural autoregressive flows from which it is easy to sample. We leverage these models in quantile flow regression, an approach that parameterizes predictive conditional distributions with flows, resulting in improved probabilistic predictions on tasks such as time series forecasting and object detection. Our novel objective functions and neural flow parameterizations also yield improvements on popular generation and density estimation tasks, and represent a step beyond maximum likelihood learning of flows.