NCPU-EM-1D

A one-dimensional neural cellular automaton. One small local rule, applied to every cell over and over, transforms a binary string. A learned program selects the task.

Loading model…

Input

Result loading model…

Target
Output
Raw tape
Settling

Channel over time

Rows are time steps, time runs down; columns are tape cells. Click or drag to pick a step. On the I/O channel, −1 reads as 0, +1 as 1, and values near 0 as blank.

 

All channels at step 0

 

Decoded at this step:

What the tasks do

The input is x = x1x2…xn, with x1 the leftmost bit; y is the output and ⊕ is XOR. Examples use the input .