Skip to content

Application of the LIME algorithm by Marco Tulio Ribeiro, Sameer Singh, Carlos Guestrin to the domain of time series classification

Notifications You must be signed in to change notification settings

emanuel-metzenthin/Lime-For-Time

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

44 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

LIME For Time

We applied the LIME algorithm (LIME - Local Interpretable Model-Agnostic Explanations) developed by Marco Tulio Ribeiro, Sameer Singh and Carlos Guestrin (paper, GitHub) to time series classification.

LIME is used to better understand predictions made by complex black-box ML models. By making use of a transparent, interpretable model the algorithm tries to approximate the black-box locally in vicinity of a specific data sample.

A given example instance is perturbed (i.e. parts of its features are 'switched off', pixels greyed out for example) many times and fed to the black-box. The approximating model then learns which features have the most influence on the final prediction.

The LIME package only supports image, text and tabular data. We extended it to be able to deal with time series data.

Demo

See the demo notebook for how to use the module.

About

Application of the LIME algorithm by Marco Tulio Ribeiro, Sameer Singh, Carlos Guestrin to the domain of time series classification

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 4

  •  
  •  
  •  
  •  

Languages