Deep Generative Quantile-Copula Models for Probabilistic Forecasting

By Ruofeng Wen, Kari Torkkola. 2019

38 citations recorded by the source corpus.

We introduce a new category of multivariate conditional generative models and demonstrate its performance and versatility in probabilistic time series forecasting and simulation. Specifically, the output of quantile regression networks is expanded from a set of fixed quantiles to the whole Quantile Function by a univariate mapping from a latent uniform distribution to the target distribution. Then the multivariate case is solved by learning such quantile functions for each dimension's marginal distribution, followed by estimating a conditional Copula to associate these latent uniform random variables. The quantile functions and copula, together defining the joint predictive distribution, can be parameterized by a single implicit generative Deep Neural Network.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Probabilistic & Bayesian Methods, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Bryan Lim
2. Jan Gasthaus
3. Tim Januschowski
4. Bernie Wang
5. Stefan Zohren
6. Danielle C. Maddix
7. Laurent Callot
8. Syama Sundar Rangapuram

In Computer Vision and Pattern Recognition

1. Laurent Callot
2. Michael Bohlke-Schneider
3. Bryan Lim
4. David Salinas
5. Konstantinos Benidis
6. Lorenzo Stella
7. Valentin Flunkert
8. Jan Gasthaus

In Probabilistic & Bayesian Methods

1. Jan Gasthaus
2. Tim Januschowski
3. Bernie Wang
4. Danielle C. Maddix
5. David Salinas
6. Valentin Flunkert
7. Bryan Lim
8. Stefan Zohren

In Signal Processing

1. Bernie Wang
2. Bryan Lim
3. Tim Januschowski
4. Jan Gasthaus
5. Stefan Zohren
6. Syama Sundar Rangapuram
7. Danielle C. Maddix
8. Laurent Callot

Citations in the corpus

• Time-series forecasting with deep learning: a survey (2020, background): “re events – e.g. allowing risk managers to insulate portfolios against market crashes. A common way to model uncertainties is to use deep neural networks to generate parameters of known distributions [27, 37, 38]. For example, Gaussian distributions are typically used for forecasting problems with c”
• High-Dimensional Multivariate Forecasting with Low-Rank Gaussian Copula Processes (2019, methodology): “A concurrent approach was proposed in [35] which also uses Copula and estimates marginal quantile functions with the approach proposed in [11] and models the Cholesky factor as the output of a neural network.”
• Deep Learning for Time Series Forecasting: Tutorial and Literature Survey (2020, methodology): “Zhang [201] MLP Point 1 MSE MSE, MAD sunspot, lynx, exchange rate Hybrid local model that uses ARIMA to capture the linear component and a NN for the nonlinear residuals Khashei and Bijari [100] MLP Point 1 MSE MAE, MSE sunspot, lynx, exchange rate Hybrid local model that uses ARIMA and a NN for tre”