The source code and data of the paper "Instance-wise Graph-based Framework for Multivariate Time Series Forecasting".

Last update: Dec 05, 2022

Related tags

Deep Learning IGMTF

Overview

IGMTF

The source code and data of the paper "Instance-wise Graph-based Framework for Multivariate Time Series Forecasting".

Requirements

The framework is implemented using python3 with dependencies specified in requirements.txt

Running the code

git clone https://github.com/Wentao-Xu/IGMTF.git
cd IGMTF
tar -zxvf data.tar.gz
mkdir model

Traffic

# Horizon 3
python learn.py --save ./model/model-traffic-3.pt --data ./data/traffic.txt --num_nodes 862 --epoch 100 --horizon 3 --hidden_size 256 --hidden_batch_size 64 --k_day 30 --n_neighbor 20

# Horizon 6
python learn.py --save ./model/model-traffic-6.pt --data ./data/traffic.txt --num_nodes 862 --epoch 100 --horizon 6 --hidden_size 256 --hidden_batch_size 64 --k_day 5 --n_neighbor 30

# Horizon 12
python learn.py --save ./model/model-traffic-12.pt --data ./data/traffic.txt --num_nodes 862 --epoch 100 --horizon 12 --hidden_size 256 --hidden_batch_size 64 --k_day 10 --n_neighbor 30

# Horizon 24
python learn.py --save ./model/model-traffic-24.pt --data ./data/traffic.txt --num_nodes 862 --epoch 100 --horizon 24 --hidden_size 256 --hidden_batch_size 64 --k_day 3 --n_neighbor 10

Electricity

# Horizon 3
python learn.py --save ./model/model-electricity-3.pt --data ./data/electricity.txt --num_nodes 321 --epoch 100 --horizon 3 --hidden_size 512 --k_day 5 --n_neighbor 20

# Horizon 6
python learn.py --save ./model/model-electricity-6.pt --data ./data/electricity.txt --num_nodes 321 --epoch 100 --horizon 6 --hidden_size 512 --k_day 3 --n_neighbor 3

# Horizon 12
python learn.py --save ./model/model-electricity-12.pt --data ./data/electricity.txt --num_nodes 321 --epoch 100 --horizon 12 --hidden_size 512 --k_day 10 --n_neighbor 5

# Horizon 24
python learn.py --save ./model/model-electricity-24.pt --data ./data/electricity.txt --num_nodes 321 --epoch 100 --horizon 24 --hidden_size 512 --k_day 5 --n_neighbor 20

Exchange-Rate

# Horizon 3
python learn.py --save ./model/model-exchange-3.pt --data ./data/exchange_rate.txt --num_nodes 8 --epoch 100 --horizon 3 --hidden_size 512 --batch_size 16 --k_day 20 --n_neighbor 20

# Horizon 6
python learn.py --save ./model/model-exchange-6.pt --data ./data/exchange_rate.txt --num_nodes 8 --epoch 100 --horizon 6 --hidden_size 512 --batch_size 16 --k_day 5 --n_neighbor 10

# Horizon 12
python learn.py --save ./model/model-exchange-12.pt --data ./data/exchange_rate.txt --num_nodes 8 --epoch 100 --horizon 12 --hidden_size 512 --batch_size 16 --k_day 10 --n_neighbor 10

# Horizon 24
python learn.py --save ./model/model-exchange-24.pt --data ./data/exchange_rate.txt --num_nodes 8 --epoch 100 --horizon 24 --hidden_size 512 --batch_size 16 --k_day 5 --n_neighbor 20

The source code and data of the paper "Instance-wise Graph-based Framework for Multivariate Time Series Forecasting".

Related tags

Overview

IGMTF

Requirements

Running the code

Owner

Wentao Xu

PyTorch code for Composing Partial Differential Equations with Physics-Aware Neural Networks

Machine Learning University: Accelerated Computer Vision Class

Material for my PyConDE & PyData Berlin 2022 Talk "5 Steps to Speed Up Your Data-Analysis on a Single Core"

KakaoBrain KoGPT (Korean Generative Pre-trained Transformer)

A Human-in-the-Loop workflow for creating HD images from text

Data-depth-inference - Data depth inference with python

The aim of this project is to build an AI bot that can play the Wordle game, or more generally Squabble

Revisiting Self-Training for Few-Shot Learning of Language Model.

Global-Local Path Networks for Monocular Depth Estimation with Vertical CutDepth [Paper]

QHack—the quantum machine learning hackathon

Model parallel transformers in Jax and Haiku

Speech-Emotion-Analyzer - The neural network model is capable of detecting five different male/female emotions from audio speeches. (Deep Learning, NLP, Python)

This is the code for the paper "Motion-Focused Contrastive Learning of Video Representations" (ICCV'21).

TransMVSNet: Global Context-aware Multi-view Stereo Network with Transformers.

JAX code for the paper "Control-Oriented Model-Based Reinforcement Learning with Implicit Differentiation"

Direct design of biquad filter cascades with deep learning by sampling random polynomials.

Code for a seq2seq architecture with Bahdanau attention designed to map stereotactic EEG data from human brains to spectrograms, using the PyTorch Lightning.

Anti-Adversarially Manipulated Attributions for Weakly and Semi-Supervised Semantic Segmentation (CVPR 2021)

An Exact Solver for Semi-supervised Minimum Sum-of-Squares Clustering