DiffLogic CA with 128-bit cell state and 12 steps successfully learns 20x20 lizard pattern, generalizing to 40x40 grid

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• DiffLogic CA with 64 channels successfully learns 16x16 colored Google G pattern in 15 steps using 8-color binary palettefinding0.867
  Demonstration of multi-channel RGB color pattern generation with binary states
• DiffLogic CA fully converges to 16x16 checkerboard target pattern with both soft and hard losses reaching zerofinding0.838
  Core result of pattern generation experiment demonstrating recurrent circuit learning
• DiffLogic CA has strong kinship with CAM-8 architecture where each cell uses DRAM for state variables and SRAM for processingclaim0.815
  Authors' architectural analogy between DiffLogic CA and Toffoli-Margolus CAM-8
• DiffLogic CA can directly learn the local rules needed to achieve desired macroscopic computation, addressing the fundamental challenge identified by Toffoli and Margolusclaim0.815
  Core thesis of the paper framed against the historical challenge of hand-crafting CA rules
• Asynchronously trained DiffLogic CA shows greater robustness to 10x10 pixel damage than synchronously trained version, measured by sum of absolute differencesfinding0.800
  Quantitative comparison of synchronous vs asynchronous training for noise resilience
• Asynchronous DiffLogic CA requires 50 steps vs 20 steps for synchronous training to reconstruct checkerboard patternfinding0.795
  Cost of asynchronous training in terms of convergence time steps
• Synchronously trained DiffLogic CA circuit succeeds at asynchronous inference without retrainingfinding0.778
  Unexpected result demonstrating robustness of learned circuits beyond their training regime
• Checkerboard circuit trained on 16x16 grid successfully generalizes to 64x64 grid with 4x more time stepsfinding0.765
  Grid scaling generalization result demonstrating boundary-size invariance