Photon mapping
A two-pass global-illumination method: shoot photons from the lights, store where they land, then read that map back when rendering the image.
Henrik Wann Jensen introduced the technique in 1996. The first pass traces photons from the light sources through the scene and stores their hits in a photon map, usually a kd-tree. The second pass renders normally but estimates the indirect lighting by gathering nearby photons around each point. The method is biased but converges stably.
Its strength is exactly what pure ray tracing and path tracing struggle with: caustics (the shimmering pattern on the bottom of a pool), color bleeding and light in haze and smoke. Unlike radiosity, which only handles diffuse surfaces, the photon map also deals with specular and transparent materials. It is mostly used offline, but ideas from it live on in real-time GI approximations.