Skip to content

Multi-directional Recurrent Neural Networks (MRNN) - IEEE TBME 2019

Notifications You must be signed in to change notification settings

jsyoon0823/MRNN

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

25 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Codebase for "Estimating Missing Data in Temporal Data Streams Using Multi-Directional Recurrent Neural Networkss (MRNN)"

Authors: Jinsung Yoon, William R. Zame, Mihaela van der Schaar

Paper: Jinsung Yoon, William R. Zame, Mihaela van der Schaar, "Estimating Missing Data in Temporal Data Streams Using Multi-Directional Recurrent Neural Networks," IEEE Transactions on Biomedical Engineering, 2019.

Paper Link: https://ieeexplore.ieee.org/document/8485748

Contact: [email protected]

This directory contains implementations of MRNN framework for imputation in time-series data using GOOGLE stocks dataset.

To run the pipeline for training and evaluation on MRNN framwork, simply run python3 -m main_mrnn.py.

Command inputs:

  • file_name: data file name
  • seq_len: sequence length of time-series data
  • miss_rate: probability of missing components (to be introduced)
  • h_dim: hidden state dimensions
  • batch_size: the number of samples in mini batch
  • iteration: the number of iteration
  • learning_rate: learning rate of model training
  • metric_name: imputation performance metric

Example command

$ python3 main_mrnn.py --file_name data/google.csv --seq_len 7 
--missing_rate: 0.2 --h_dim 10 --batch_size 128 --iteration 2000
--learning_rate 0.01 --metric_name rmse

Outputs

  • x: original data with missing
  • ori_x: original data without missing
  • m: mask matrix
  • t: time matrix
  • imputed_x: imputed data
  • performance: imputation performance

About

Multi-directional Recurrent Neural Networks (MRNN) - IEEE TBME 2019

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages