TATD

This is a PyTorch and TensorLy implementation of Time-Aware Tensor Decomposition for Sparse Tensors.

Prerequisites

• Python 3.6+
• NumPy
• pandas
• scikit-learn
• PyTorch
• TensorLy

Usage

• Install all of the prerequisites
• You can run the demo script by bash demo.sh, which simply runs src/main.py.
• You can change the datasets and hyper-parameters by modifying src/main.py.
• you can check out the running results in out directory.

Datasets

Preprocessed datasets are in the data directory.

Name	Description	Size	NNZ	Granularity in Time	Original Source
Beijing Air Quality	time x locations x pollutants	35064 x 12 x 6	2454305	hourly	Link
Madrid Air Quality	time x locations x pollutants	2678 x 26 x 17	337759	daily	Link
Radar Traffic	time x locations x directions	17937 x 17 x 5	495685	hourly	Link
Indoor Condition	time x locations x sensor	19735 x 9 x 2	241201	every 10 minutes	Link
Server Room	time x air conditioning x server power x locations	4157 x 3 x 3 x 34	1009426	1 second	Link