Neural Networks Learn to Walk

can a neural network learn to walk as a physical object in a physics simulation? here I train walking neural nets with an evolutionary algorithm. The input nodes/feet are activated by sine waves at learned phases & connections between two neurons extend based on their difference generation 1, batch 5 generation 2, batch 4 generation 2, batch 1 generation 18, batch 1 generation 18, batch 4 generation 19, batch 3 best at generation 20

A tweet shows an evolutionary algorithm training groups of neural-network creatures to walk in a physics simulation. Across the clip, many tangled networks move and fail until only one small structure remains as the best result at generation 20. The joke is the dramatic contrast between sophisticated technical language and the underwhelming final walking machine, making it useful as a reaction to arduous optimization that produces a barely functional result. EN: A tweet shows an evolutionary algorithm training groups of neural-network creatures to walk in a physics simulation. Across the clip, many tangled networks move and fail until only one small structure remains as the best result at generation 20. The joke is the dramatic contrast between sophisticated technical language and the underwhelming final walking machine, making it useful as a reaction to arduous optimization that produces a barely functional result.

• @ithueti / 9244 2023-04-12