PyTorch implementation of "One-shot Pruning of Gated Recurrent Unit Neural Network by Sensitivity for Time-series Prediction" by Hong Tang, Xiangzheng Ling, Liangzhi Li (Member, IEEE) et al.

Following packages are required for this project

• Python 3.6+

• PyTorch-GPU 0.4.1

• tqdm, csv, time

• numpy, pandas

• matplotlib

• argparse

1. train a simple local areal network traffic

python main.py --k_level 0.01 --sensitivity 3.754

2. training comparison model GVGRUs

python GVGRU.py --model [Model_Name]

1. The test results of the standard GRU (Baseline).

2. The test results of SCGRU (Our).

1. The test results of the standard GRU (Baseline).

2. The test results of SCGRU (Our).

Note that the power load data set used in the paper is temporarily unavailable due to its industry sensitivity. Please understand.