search

marcotcr / lime

Lime: Explaining the predictions of any machine learning classifier
BSD 2-Clause "Simplified" License
11.5k stars 1.79k forks source link

Selection of Kernel width #587

Closed dsvrsec closed 3 years ago

dsvrsec commented 3 years ago

Hi, Can you please suggest on what should be the kernel width for a specific image and how to derive it.Thank you

CarlosGuemS commented 3 years ago

Hi, maybe this is answer is unsatisfactory to you, but seeing nobody is answering you...

The issue of the kernel width is an open one and there is no easy solution. There are some methods like OptiLIME that focuses in how obtaining the kernel width for a given problem. I'll link you the paper, and I hope it is helpful: https://arxiv.org/abs/2006.05714

marcotcr commented 3 years ago

see #222