Adjusting for Autocorrelated Errors in Neural Networks for Time Series

Last update: Nov 5, 2022

Related tags

Overview

Adjusting for Autocorrelated Errors in Neural Networks for Time Series

This repository is the official implementation of the paper "Adjusting for Autocorrelated Errors in Neural Networks for Time Series" (arXiv link).

For simplicity, we only provide the code for time series forecasting. However, it is straightforward to implement our method with other models on other time series tasks as described in the appendix of the paper.

Requirements

To install requirements:

pip3 install -r requirements.txt

Datasets

Available datasets are located in the directory data/forecasting. Traffic is not included because it exceeds the 100MB size limit set by Github. However, you can download it here and format the it into a .npy file. ADI-related datasets are not released because they are proprietary.

To use your own dataset, format it into a numpy array with size TxN and saved it into the data directory as a .npy file.

Training and Evaluation

Example commands can be found in run.sh.

Citation

@inproceedings{sun2021adjusting,
    title={Adjusting for Autocorrelated Errors in Neural Networks for Time Series Regression and Forecasting}, 
    author={Fan-Keng Sun and Christopher I. Lang and Duane S. Boning},
    booktitle = {Advances in Neural Information Processing Systems},
    year={2021},
}

Comments

some question about 𝐗_𝑡−𝜌𝐗_(𝑡−1) in the code

Hello! as the paper said，the input and output should be： but the code in forecasting_runner: if args.inp_adj: x = torch.cat([avg[None].repeat(bs, 1, 1), x], dim=1) x = x[:, 1:] - rho * x[:, :-1] prd_y = self.model(x) if args.out_adj: prd_y += rho * x[:, -1] the x[:, -1] under has changed which mean that x[:, -1]=x[:, 1:] - rho * x[:, :-1] instead of x_{t-1}. Do I misunderstand?

opened by Lanturewen 1
i have some questions about forecasting_runner.py

Hey, bro, I don't quite understand the error correction of the input and output of the prediction part in the code. Why should the average value be added and subtracted in the input correction, and why should the Rho multiplied by the last column of X be added in the output correction? These things don't seem to be reflected in the paper. Can you help me solve them?My email is [email protected]

opened by pharaohback 1

This repository contains notebook implementations of the following Neural Process variants: Conditional Neural Processes (CNPs), Neural Processes (NPs), Attentive Neural Processes (ANPs).

The Neural Process Family This repository contains notebook implementations of the following Neural Process variants: Conditional Neural Processes (CN

892 Dec 28, 2022

TCNN Temporal convolutional neural network for real-time speech enhancement in the time domain

TCNN Pandey A, Wang D L. TCNN: Temporal convolutional neural network for real-time speech enhancement in the time domain[C]//ICASSP 2019-2019 IEEE Int

16 Dec 30, 2022

[CVPR 2022] Official code for the paper: "A Stitch in Time Saves Nine: A Train-Time Regularizing Loss for Improved Neural Network Calibration"

MDCA Calibration This is the official PyTorch implementation for the paper: "A Stitch in Time Saves Nine: A Train-Time Regularizing Loss for Improved

21 Dec 22, 2022

tsai is an open-source deep learning package built on top of Pytorch & fastai focused on state-of-the-art techniques for time series classification, regression and forecasting.

Time series Timeseries Deep Learning Pytorch fastai - State-of-the-art Deep Learning with Time Series and Sequences in Pytorch / fastai

2.8k Jan 8, 2023

Adjusting for Autocorrelated Errors in Neural Networks for Time Series

Related tags

Overview

Adjusting for Autocorrelated Errors in Neural Networks for Time Series

Requirements

Datasets

Training and Evaluation

Citation

You might also like...

This repository contains notebook implementations of the following Neural Process variants: Conditional Neural Processes (CNPs), Neural Processes (NPs), Attentive Neural Processes (ANPs).

TCNN Temporal convolutional neural network for real-time speech enhancement in the time domain

[CVPR 2022] Official code for the paper: "A Stitch in Time Saves Nine: A Train-Time Regularizing Loss for Improved Neural Network Calibration"

A unified framework for machine learning with time series

Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting

Implementation of the paper NAST: Non-Autoregressive Spatial-Temporal Transformer for Time Series Forecasting.

MINIROCKET: A Very Fast (Almost) Deterministic Transform for Time Series Classification

AntroPy: entropy and complexity of (EEG) time-series in Python

tsai is an open-source deep learning package built on top of Pytorch & fastai focused on state-of-the-art techniques for time series classification, regression and forecasting.

Comments

some question about 𝐗_𝑡−𝜌𝐗_(𝑡−1) in the code

i have some questions about forecasting_runner.py

Owner

Fan-Keng Sun

This project is a loose implementation of paper "Algorithmic Financial Trading with Deep Convolutional Neural Networks: Time Series to Image Conversion Approach"

Delving into Localization Errors for Monocular 3D Object Detection, CVPR'2021

DeepConsensus uses gap-aware sequence transformers to correct errors in Pacific Biosciences (PacBio) Circular Consensus Sequencing (CCS) data.

Alleviating Over-segmentation Errors by Detecting Action Boundaries

Deep learning-based approach to discovering Granger causality networks in multivariate time series

Code for the paper "TadGAN: Time Series Anomaly Detection Using Generative Adversarial Networks"

Library for implementing reservoir computing models (echo state networks) for multivariate time series classification and clustering.

Spectral Temporal Graph Neural Network (StemGNN in short) for Multivariate Time-series Forecasting

A real world application of a Recurrent Neural Network on a binary classification of time series data

Complex-Valued Neural Networks (CVNN)Complex-Valued Neural Networks (CVNN)