Life, recursed randomly perturbs a single entry in the kernel after each generation. This simulates a recursive 'branching' from end state to start state, where each generation makes a non-deterministic 'leap' to a parent generation. Each parent and child generation can be thought of as a pair of start and end states, where the superset of pairs is loosely connected by inter-generation perturbations, spanning the combinatorially huge set of all initial states. View archive.
Life, recursed
	Life, stochastic adds an element of chance to each pass of the kernel. Random probabilities are generated for each rule type, and these probabilities guide the application of the kernel on a per-pixel basis. For example, a kernel consisting of a single 'invert' rule may only have a 75% chance of being applied to each pixel. View archive.
Life, stochastic
	Life, invertible generates a random kernel, which is inverted after a number of generations, allowing the simulation to revert from the start state back to the end state. Not all kernels are invertible, due to inherent information loss in the logical 'or' and 'and' operations used in 'on' and 'off' rules. '<input> or 1' discards whether one or both inputs were 1. '<input> and 0' discards whether one or both inputs were 0. Because of this loss, the kernel can only be composed of 'null' (leave alone) and 'invert' rules.
Life, invertible
	Conway's Life does not have an explicit kernel, but uses kernel-based methods to count the number of neighbors any pixel has. An 'on' pixel with 1, 4, or more neighbors will die. An 'off' pixel with exactly 3 neighbors will come to life. The pixel state is untouched in all other configurations.
Conway's Life
Eidolon, 2003.