4 ms·
Show HN: Can an AI model fit on a single pixel?
- deevelton 6mo agoThis was just a fun thought experiment. Feed it two inputs (e.g. chance of rain and wind speed; this is one of the examples in the demo) and it learns to answer a yes/no question like "bring an umbrella?" It's a one-neuron binary classifier with three learned parameters: two weights and a bias. Those three numbers map directly to Red, Green, and Blue. Save the model: you get a 1x1 PNG. Load the pixel: you get your classifier back. The color is the model.
- mememememememo 6mo agoIt is fun. The trainer seems to get stuck in local minima though.
- deevelton 6mo agoThanks; does this fix it? https://github.com/dvelton/ai-pixel/commit/ba04e1c649e0ec99b7e4ea38807ba96002139b33 https://github.com/dvelton/ai-pixel/commit/ba04e1c649e0ec99b...
- mememememememo 6mo agoTried again and no. To reproduce create a corridor e.g. X O X O X O X O X O Line can get jammed in non vertical positon.
- deevelton 6mo agoIncreased default epochs to 2000 instead of 500, I think that fixed it but let me know if you still run into this.
- mememememememo 6mo agoWorks for me now, thanks!
- deevelton 6mo agoawesome! thanks for checking it out
- crackalamoo 6mo agoThis is super fun! Very elegant idea Has anyone done this for larger neural nets? Is there a way to extract some kind of pattern or is the image just noise no matter how you construct it? I'd be curious to see something like that
- deevelton 6mo agoNot sure what happens if you start scaling this, but curious too. Would love to see someone who actually knows ML try it.
- deevelton 6mo agoone of my teammates experimenting with something along these lines :) https://github.com/jessephus/model-mosaic https://github.com/jessephus/model-mosaic