71 Hits in 3.4 sec

Probabilistic Forecasting with Spline Quantile Function RNNs

Jan Gasthaus, Konstantinos Benidis, Yuyang Wang, Syama Sundar Rangapuram, David Salinas, Valentin Flunkert, Tim Januschowski
2019 International Conference on Artificial Intelligence and Statistics  
In this paper, we propose a flexible method for probabilistic modeling with conditional quantile functions using monotonic regression splines.  ...  Within this framework, we propose a method for probabilistic time series forecasting, which combines the modeling capacity of recurrent neural networks with the flexibility of a splinebased representation  ...  FORECASTING WITH SQF-RNN We now turn to how such a spline-based representation of the quantile function can be combined with recurrent neural networks and applied to probabilistic forecasting, yielding  ... 
dblp:conf/aistats/GasthausBWRSFJ19 fatcat:6pjwlvyx2zcptbcbriuvr5shzm

Probabilistic Time Series Forecasting with Implicit Quantile Networks [article]

Adèle Gouttes, Kashif Rasul, Mateusz Koren, Johannes Stephan, Tofigh Naghibi
2021 arXiv   pre-print
Here, we propose a general method for probabilistic time series forecasting.  ...  We combine an autoregressive recurrent neural network to model temporal dynamics with Implicit Quantile Networks to learn a large class of distributions over a time-series target.  ...  Our approach is closely related to the SQF-RNN (Gasthaus et al., 2019) which models the conditional quantile function using isotonic splines.  ... 
arXiv:2107.03743v1 fatcat:wa2ls7hf3zdg3bpmwgb6bxbrii

Deep Learning for Time Series Forecasting: Tutorial and Literature Survey

Konstantinos Benidis, Syama Sundar Rangapuram, Valentin Flunkert, Yuyang Wang, Danielle Maddix, Caner Turkmen, Jan Gasthaus, Michael Bohlke-Schneider, David Salinas, Lorenzo Stella, François-Xavier Aubet, Laurent Callot (+1 others)
2022 ACM Computing Surveys  
the recent deep forecasting literature.  ...  Deep learning based forecasting methods have become the methods of choice in many applications of time series prediction or forecasting often outperforming other approaches.  ...  Interestingly, a popular discretization strategy, adaptive binning, used with multinomial distributions corresponds to quantile functions parametrized by piece-wise linear splines, see Fig. 4 .  ... 
doi:10.1145/3533382 fatcat:l46f34dbp5fdpawbpnoiippf6q

Day-ahead nonparametric probabilistic forecasting of photovoltaic power generation based on the LSTM-QRA ensemble model

Fei. Mei, Jiaqi. Gu, Jixiang. Lu, Jinjun. Lu, Jiatang. Zhang, Yuhan. Jiang, Tian. Shi, Jianyong. Zheng
2020 IEEE Access  
Quantile regression averaging (QRA) is used to ensemble a group of independent long short-term memory (LSTM) deterministic forecasting models for obtaining the probabilistic forecasting of PV output.  ...  Additionally, in comparison with the benchmark methods, LSTM-QRA has higher prediction performance due to the better forecasting accuracy of independent deterministic forecasts.  ...  The feasibility of generating a nonparametric probabilistic forecasting model by integrating independent deterministic models with QRA is verified.  The relationship between the pinball loss function  ... 
doi:10.1109/access.2020.3021581 fatcat:h3etgfpsqrai7au7ct3qscfj64

GluonTS: Probabilistic and Neural Time Series Modeling in Python

Alexander Alexandrov, Konstantinos Benidis, Michael Bohlke-Schneider, Valentin Flunkert, Jan Gasthaus, Tim Januschowski, Danielle C. Maddix, Syama Sundar Rangapuram, David Salinas, Jasper Schulz, Lorenzo Stella, Ali Caner Türkmen (+1 others)
2020 Journal of machine learning research  
We introduce the Gluon Time Series Toolkit (GluonTS), a Python library for deep learning based time series modeling for ubiquitous tasks, such as forecasting and anomaly detection.  ...  DeepAR uses a recurrent neural network (RNN) with LSTM or GRU cells, and estimates parameters of a parametric distribution or uses a parameterization of the quantile function (Flunkert et al., 2019; .  ...  MQ-RNN and MQ-CNN combine RNN and dilatied causal convolution (CNN) encoders, respectively, with a quantile decoder (Wen et al., 2017) , using the flexible sequence-to-sequence framework provided in GluonTS  ... 
dblp:journals/jmlr/0003BBFGJMRSSST20 fatcat:kdscp3udd5ebpbtlyrynni5wdm

Probabilistic Forecasting with Temporal Convolutional Neural Network [article]

Yitian Chen, Yanfei Kang, Yixiong Chen, Zizhuo Wang
2020 arXiv   pre-print
We present a probabilistic forecasting framework based on convolutional neural network for multiple related time series forecasting.  ...  Combined with representation learning, our approach is able to learn complex patterns such as seasonality, holiday effects within and across series, and to leverage those patterns for more accurate forecasts  ...  More recently, Gasthaus et al. (2019) propose SQF-RNN, a probabilistic framework to model conditional quantile functions with isotonic splines, which allows more flexible output distributions.  ... 
arXiv:1906.04397v3 fatcat:g4lwo4psbnbqflnhc6frqzhfea

Learning Quantile Functions without Quantile Crossing for Distribution-free Time Series Forecasting [article]

Youngsuk Park, Danielle Maddix, François-Xavier Aubet, Kelvin Kan, Jan Gasthaus, Yuyang Wang
2022 arXiv   pre-print
In this work, we propose the Incremental (Spline) Quantile Functions I(S)QF, a flexible and efficient distribution-free quantile estimation framework that resolves quantile crossing with a simple neural  ...  Equipped with the analytical evaluation of the continuous ranked probability score of I(S)QF representations, we apply our methods to NN-based times series forecasting cases, where the savings of the expensive  ...  As an example, DeepAR, a RNN-based probabilistic forecaster, offers likelihood choices of normal, student-t, negative binomial, etc.  ... 
arXiv:2111.06581v2 fatcat:dpd663a3gvajhoqlmvlj6twaym

Quantifying Uncertainty in Deep Spatiotemporal Forecasting [article]

Dongxia Wu, Liyao Gao, Xinyue Xiong, Matteo Chinazzi, Alessandro Vespignani, Yi-An Ma, Rose Yu
2021 arXiv   pre-print
In high stakes domains, being able to generate probabilistic forecasts with confidence intervals is critical to risk assessment and decision making.  ...  Computationally, quantile regression type methods are cheaper for a single confidence interval but require re-training for different intervals.  ...  Another solution to alleviate the quantile crossing issue is to minimize CRPS by assuming the quantile function to be a piecewise linear spline with monotonicity [19] , a method we call Spline Quantile  ... 
arXiv:2105.11982v2 fatcat:yjfc6ahjmjcftnvoebgope4zmi

DeepGLEAM: A hybrid mechanistic and deep learning model for COVID-19 forecasting [article]

Dongxia Wu, Liyao Gao, Xinyue Xiong, Matteo Chinazzi, Alessandro Vespignani, Yi-An Ma, Rose Yu
2021 arXiv   pre-print
We introduce DeepGLEAM, a hybrid model for COVID-19 forecasting. DeepGLEAM combines a mechanistic stochastic simulation model GLEAM with deep learning.  ...  We demonstrate DeepGLEAM on real-world COVID-19 mortality forecasting tasks.  ...  Next, we discuss probabilistic forecasts by combing DeepGLEAM with Frequentist or Bayesian uncertainty quantification (UQ) methods, including bootstrap, quantile regression (Quantile), spline quantile  ... 
arXiv:2102.06684v3 fatcat:5kh3yivg3nauvbpz2ti6mne5ta

VQ-AR: Vector Quantized Autoregressive Probabilistic Time Series Forecasting [article]

Kashif Rasul, Young-Jin Park, Max Nihlén Ramström, Kyung-Min Kim
2022 arXiv   pre-print
In practice, deep learning based time series models come in many forms, but at a high level learn some continuous representation of the past and use it to output point or probabilistic forecasts.  ...  Time series models aim for accurate predictions of the future given the past, where the forecasts are used for important downstream tasks like business decision making.  ...  [14] : an RNN based non-parametric method which models the quantiles via linear splines and also regresses the Quantile loss; • IQN-RNN [16] : combines an RNN model with an Implicit Quantile Network  ... 
arXiv:2205.15894v1 fatcat:e7kv6n22frebhdy6gaxnrgstmm

The Effectiveness of Discretization in Forecasting: An Empirical Study on Neural Time Series Models [article]

Stephan Rabanser, Tim Januschowski, Valentin Flunkert, David Salinas, Jan Gasthaus
2020 arXiv   pre-print
To remedy this, we evaluate the forecasting accuracy of instances of the aforementioned model classes when combined with different types of data scaling and binning.  ...  Time series modeling techniques based on deep learning have seen many advancements in recent years, especially in data-abundant settings and with the central aim of learning global models that can extract  ...  In these approaches, instead of modeling the entire output distribution, only a fixed set of quantile levels is predicted. The spline quantile function approach of Gasthaus et al.  ... 
arXiv:2005.10111v1 fatcat:akiig3mwcfcinci3nnssxcaepe

Applications of Probabilistic Forecasting in Smart Grids: A Review

Hosna Khajeh, Hannu Laaksonen
2022 Applied Sciences  
This paper reviews the recent studies and works dealing with probabilistic forecasting models and their applications in smart grids.  ...  Unlike the existing reviews, this paper assesses several uncertain parameters for which probabilistic forecasting models have been developed.  ...  Then, an RNN-LSTM network is trained to find the optimal values of PDF parameters. The other model is, however, a nonparametric probabilistic forecasting model that is integrated with RNN-LSTM.  ... 
doi:10.3390/app12041823 doaj:7983bd9658584546868ebcb5964433cc fatcat:ddrmqguhmbe6dpn5nbw3cl43bm

Neural forecasting: Introduction and literature overview [article]

Konstantinos Benidis, Syama Sundar Rangapuram, Valentin Flunkert, Bernie Wang, Danielle Maddix, Caner Turkmen, Jan Gasthaus, Michael Bohlke-Schneider, David Salinas, Lorenzo Stella, Laurent Callot, Tim Januschowski
2020 arXiv   pre-print
Neural network based forecasting methods have become ubiquitous in large-scale industrial forecasting applications over the last years.  ...  Building on these foundations, the article then gives an overview of the recent literature on neural networks for forecasting and applications.  ...  The spline quantile function RNN model (SQF-RNN) proposed in [62] uses the same basic RNN architecture as DeepAR, but uses a spline-based parametrization of the quantile function of the output distribution  ... 
arXiv:2004.10240v1 fatcat:i3nbsw5sojckdee5d4magpwsl4

Deep Factors for Forecasting [article]

Yuyang Wang, Alex Smola, Danielle C. Maddix, Jan Gasthaus, Dean Foster, Tim Januschowski
2019 arXiv   pre-print
Producing probabilistic forecasts for large collections of similar and/or dependent time series is a practically relevant and challenging task.  ...  Probabilistic Chung, J., Kastner, K., Dinh, L., Goel, K., Courville, A. C., forecasting with spline quantile function rnns. In The and Bengio, Y.  ...  The RNN cell for RNN Forecaster our algorithm, we compare with DeepAR (DA), a state-of- has an output dimension of 1 while that of DeepFactor is of art RNN-based probabilistic forecasting  ... 
arXiv:1905.12417v1 fatcat:6jzizdhebfb43goezaplgsonka

Learning Probability Distributions in Macroeconomics and Finance [article]

Jozef Barunik, Lubos Hanus
2022 arXiv   pre-print
We propose a deep learning approach to probabilistic forecasting of macroeconomic and financial time series.  ...  Quantile function represented by spline combined with recurrent neural network proposed by Gasthaus et al. (2019) is a distribution-free approach with objective function based on CRPS score (Gneiting  ...  and Raftery, 2007) constructed with respect to monotonicity of quantile function.  ... 
arXiv:2204.06848v1 fatcat:g3kwgjfcavd2lc2ji4y5colx7i
« Previous Showing results 1 — 15 out of 71 results