Dynamical Wasserstein Barycenters for Time Series Modeling

Dynamical Wasserstein Barycenters for Time Series Modeling

This is the code related for the Dynamical Wasserstein Barycenter model published in Neurips 2021.

To run the code and replicate the results reported in our paper,

# usage: DynamicalWassersteinBarycenters.py dataSet dataFile debugFolder interpModel [--ParamTest PARAMTEST] [--lambda LAM] [--s S]

# Sample run on MSR data                                         
>> python DynamicalWassersteinBarycenters.py MSR_Batch ../Data/MSR_Data/subj090_1.mat ../debug/MSR/subj001_1.mat Wass 

# Sample run for parameter test
>> python DynamicalWassersteinBarycenters.py MSR_Batch ../Data/MSR_Data/subj090_1.mat ../debug/ParamTest/subj001_1.mat Wass --ParamTest 1 --lambda 100 --s 1.0

The interpMethod is either Wass for the Wasserstein barycentric model or GMM` for the linear interpolation model.

Simulated Data

The simulated data and experiment included in this supplement can be replicated using using the following commands.

# Generate 2 and 3 state simulated data

>> python GenerateOptimizationExperimentData.py

>> python GenerateOptimizationExperimentData_3K.py

# usage: OptimizationExperiment.py FileIn Mode
 
 

 
To finish reading, please visit source site