N-HiTS: Neural Hierarchical Interpolation for Time Series Forecasting

By Cristian Challu, Kin G. Olivares, Boris N. Oreshkin, Federico Garza, Max Mergenthaler-Canseco, Artur Dubrawski. AAAI Conference on Artificial Intelligence, 2022

471 citations recorded by the source corpus.

Recent progress in neural forecasting accelerated improvements in the performance of large-scale forecasting systems. Yet, long-horizon forecasting remains a very difficult task. Two common challenges afflicting the task are the volatility of the predictions and their computational complexity. We introduce NHITS, a model which addresses both challenges by incorporating novel hierarchical interpolation and multi-rate data sampling techniques. These techniques enable the proposed method to assemble its predictions sequentially, emphasizing components with different frequencies and scales while decomposing the input signal and synthesizing the forecast. We prove that the hierarchical interpolation technique can efficiently approximate arbitrarily long horizons in the presence of smoothness. Additionally, we conduct extensive large-scale dataset experiments from the long-horizon forecasting literature, demonstrating the advantages of our method over the state-of-the-art methods, where NHITS provides an average accuracy improvement of almost 20% over the latest Transformer architectures while reducing the computation time by an order of magnitude (50 times). Our code is available at https://github.com/Nixtla/neuralforecast.

Subfields: Electrical and Electronic Engineering, Management Science and Operations Research, Signal Processing

View Paper: https://www.semanticscholar.org/paper/3711a745537f2f2e139ec759d36f31946fa549fd

How researchers see this paper

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

In Electrical and Electronic Engineering

1. Liang Sun
2. Tian Zhou
3. Qiang Xu
4. Ailing Zeng
5. Xue Wang
6. Rong Jin
7. Haixu Wu
8. Qingsong Wen

In Management Science and Operations Research

1. Qiang Xu
2. Ailing Zeng
3. Liang Sun
4. Tian Zhou
5. Xue Wang
6. Rong Jin
7. Shiyu Wang
8. Haixu Wu

In Signal Processing

1. Qiang Xu
2. Ailing Zeng
3. Shiyu Wang
4. Lintao Ma
5. Haixu Wu
6. Zhijian Xu
7. Tian Zhou
8. Liang Sun

Citations in the corpus

• Frequency-domain MLPs are More Effective Learners in Time Series Forecasting (2023, methodology): “N-HiTS [34] combines multi-rate input sampling and hierarchical interpolation with MLPs for univariate forecasting.”
• Non-stationary Transformers: Exploring the Stationarity in Time Series Forecasting (2022, methodology): “For univariate forecasting, we include seven competitive baselines: N-HiTS [9], N-BEATS [27], Autoformer [37], Pyraformer [23], Informer [39], Reformer [19] and ARIMA [7].”
• One Fits All: Power General Time Series Analysis by Pretrained LM (2023, methodology): “Additionally, with the bloom of deep neural networks, recurrent neural networks (RNNs), such as LSTM Hochreiter & Schmidhuber (1997) and GRU Chung et al. (2014), were designed for sequential tasks. Yet the recurrent model is inefficient for training and long-term dependencies are still under resolve”
• Large Language Models Are Zero-Shot Time Series Forecasters (2023, methodology): “For Darts, we use several meth-ods that are implemented directly in the package, including neural network models (TCN [28], N-BEATS [37], N-HiTS [11]) and simple moving average models (ARIMA [8]).”
• FITS: Modeling Time Series with 10k Parameters (2023, methodology): “It is worth highlighting that FITS requires significantly fewer parameters compared to the next smallest model, Dlinear, which has 139.7K parameters.”
• TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysis (2022, methodology): “Technically, MLP-based methods (Oreshkin et al., 2019; Challu et al., 2022; Zeng et al., 2023; Zhang et al., 2022) adopt the MLP along the temporal dimension and encode the temporal dependencies into the fixed parameter of MLP layers.”
• TimeMixer: Decomposable Multiscale Mixing for Time Series Forecasting (2024, background): “…al., 2023a; Hewage et al., 2020), RNN (Lai et al., 2018; Qin et al., 2017; Salinas et al., 2020), Transformer (Vaswani et al., 2017; Zhou et al., 2021; Wu et al., 2021; Zhou et al., 2022b; Nie et al., 2023) and MLP (Zeng et al., 2023; Zhang et al., 2022; Oreshkin et al., 2019; Challu et al., 2023).”
• FiLM: Frequency improved Legendre Memory Model for Long-term Time Series Forecasting (2022, result): “At last, (in Appendix J), though failing short in training speed compared to some MLP based models like N-HiTSChallu et al. (2022), FiLM is an efficient model with shorter per-step training time and smaller memory usage compared to baseline models in univariate forecasting tasks.”