Concept
Saliency map
A saliency map is a per-input heatmap showing how much each element of an input (or intermediate tensor) influences a model's output. Think of it like a partial derivative surface: for a scalar output y and input x, the simplest saliency map is ∂y/∂x, one gradient…
The rest of “Saliency map” is a premium feature: every concept in the library gets a precise, practitioner-focused write-up like this one, cross-linked straight from the paper summaries that use it.
Log in to unlock→