HackerTrans
TopNewTrendsCommentsPastAskShowJobs

esychology

no profile record

Submissions

Show HN: Neural Particle Automata

selforg-npa.github.io
88 points·by esychology·hace 21 días·19 comments

Show HN: High-Res Neural Cellular Automata

cells2pixels.github.io
208 points·by esychology·hace 27 días·54 comments

comments

esychology
·hace 20 días·discuss
In normal NCA cells are pixels and they can perceive their neighboring pixels (cells can't move). In NPA cells are particles and they can perceive all particles in a support radius around them and these particles can move freely. Does this answer your question?
esychology
·hace 20 días·discuss
Indeed! The system has good regeneration capabilities but it certainly has limits.

The particles can only grow reliably if they start from the egg-like initial condition. If we switch the rules mid rollout, we would get a messed up morphology.
esychology
·hace 20 días·discuss
I uploaded the videos here: https://drive.google.com/drive/folders/1V8XFzq2VkZXKG7Tw8ICv...
esychology
·hace 20 días·discuss
I really loved the distill articles. Too bad it was not continued anymore...
esychology
·hace 21 días·discuss
Thanks! Yeah I think it should be possible though it requires making the cell division/splitting a differentiable operation. But nontheless, this is indeed a very interesting and promising direction to pursue.
esychology
·hace 26 días·discuss
One NN per pattern/image (instance based training).
esychology
·hace 26 días·discuss
Checkout the Isotropic NCA blog from the Google Zurich team: https://google-research.github.io/self-organising-systems/is...
esychology
·hace 26 días·discuss
Thank you for the kind comment! Please reach out, I'm happy to have a chat.
esychology
·hace 26 días·discuss
That's not possible in the current demo but this sounds like an interesting feature to work on and add!
esychology
·hace 26 días·discuss
Texture sampling retrieves pixels by coordinate, while NCA grows them from local rules with no global lookup. The weights are actually ~3× smaller than JPEG-compressed texture maps, so it's not just memorizing the image either. The more interesting part is that it's a dynamic process, you can damage the pattern while it forms and it repairs itself.
esychology
·hace 26 días·discuss
This is awesome! The strength of the flow/advection is a bit too high imo. Maybe increase some viscosity parameter?
esychology
·hace 26 días·discuss
I think performance is not the only issue for scaling to larger grids. CUDA Convolution implementation already utilizes coalescing to improve performance. The main bottleneck is that in larger grids, cells are further apart, and it takes more steps for them to be able to communicate.
esychology
·hace 26 días·discuss
The input to the NN is just the 3x3 neighborhood around a cell. We can overlap two NNs on the same grid (through interpolation). Checkout https://meshnca.github.io to see the effect. When the brush is in graft mode, it basically allows you to paint some regions that will follow a different NN rule.
esychology
·hace 26 días·discuss
If you're familiar with CAs (e.g. Conway's Game of Life), you can think of a NeuralCA as a CA where the update rule is given by a neural network. Here we optimize the neural net weights so that it behaves a certain way (e.g. grow a lizard from a single seed).
esychology
·hace 26 días·discuss
The NeuralCA both generates and maintains the pattern. Because the NCA was not exposed to damage or erasure during training, its regeneration capability is a purely emergent phenomenon. However, this ability remains somewhat brittle, particularly when the central regions of the pattern are erased.
esychology
·hace 26 días·discuss
yeah normally NCAs have a sense of up and left. There are some isotropic variants that make the perception fully rotation-invariant.
esychology
·hace 26 días·discuss
haha yes, also the same with the worm