Get the latest tech news
Solving mazes with neural cellular automata (2021)
Interactive demonstration of Neural Cellular Maze Solver. This cellular automaton is trained to output the shortest path between the two endpoints.
In this paper, we challenged the maze solver task, in which the cells find the shortest path between two given points, like mouldy computer. This state copy techniques also has a similar effect to the sample pool based strategy described in Growing Neural Cellular Automata. In this article, as a new example of artificial swarm intelligence, we show that the Neural Cellular Automata can solve the maze solver task almost perfectly.
Or read this on Hacker News