[hardmath123]

See also, a post from mid-2020 that does something similar with a "softened" Life: http://hardmath123.github.io/conways-gradient.html

[montebicyclelo]

That's a really nice write up. It's insane how similar our approaches are.

Could it be a case of [1] (but on a non grand scale) :P? I can list my sources of inspiration: [2] [3] [4].

I also tried training convolutional networks, using the soft life set-up, but failed to get them to converge.

[1] https://en.wikipedia.org/wiki/Multiple_discovery

[2] https://kevingal.com/blog/mona-lisa-gol.html

[3] https://arxiv.org/abs/1910.00935

[4] https://nicholasrui.com/2017/12/18/convolutions-and-the-game...

[UncleOxidant]

> I also tried training convolutional networks, using the soft life set-up, but failed to get them to converge.

Do you have any idea why that might be? It seems like convolution would be a natural for this problem.

[montebicyclelo]

I didn't work on it long enough to be able to draw any conclusions, but I can speculate.

I had the gradients going through the soft life approximation (i.e. it was part of the model), rather than simply training a normal cnn with life boards as the inputs and outputs. But I think the approximation may not have good enough gradient signals.