Can be Understood, Constructed, and Leveraged (for Computation)
Generate "Emergent" Complexity and Structure
Generate Complexity Yet Are Somewhat Predictable
We Want the Conv. Kernels that predominantly learn on interior cells to also work on edge cells.
So our Cellular Automata is on a Torus to eliminate edge cases, literally.
But this means we need to create bespoke Toroidal Padding for TensorFlow.js
Fully-Connected Network Diagram
Convolutional Neural Network Diagram
Network Construction Details
Hypothesis: A Reinforcement Learning Agent Can Learn To Intelligently Interact With Some Systems Based on Simple Rules ✨