How image explanation method works?

[amindadgar]

Hi I've been reading the source code of the LIME method, but I didn't understand precisely how a ridge regression classifier is used to classify images. To demonstrate my question precisely, I've written what I've understood below:

• To train a Ridge classifier a map of zero and one is created.
• The zeros and ones represent the presence and absence of a segment of the pixel.
• The Ridge classifier is trained on these zeros and ones (Am I right about this?)
• The main model is being used just to predict the label of the exact image and its neighborhoods (Again does this is right?)

So with my understanding, If the Ridge classifier is just trained on the presence and absence of the image segments, then LIME method is not a very good method to explain the model's behavior And other alternatives such as LRP could be a better option (Because of local fidelity is hard to achieve). Did I conclude right?