The boundary of neural network trainability is fractal · HackerPapers discussions

The boundary of neural network trainability is fractal — mentioned by somethingsome

Mon, 27 Jan 2025 09:04:07 GMT

> Convolution Nice idea, but I'm also planning a book on advanced cuda programming, and I will cover convolutions there, I'll think if it is interesting to have some kind of cross chapter that is valid for both books. > Noises I cover those in normal mapping > anisotropy I have a full chapter on brdf that includes anysotropy for raster and raytracing, I use it also to derive anisotropic filtering > Color spaces I have to make choices, and I don't think I'll cover in depth color spaces, it's a topic I know, but I find it very annoying to write about, in particular, making it fun to read is…

Paper and discussion history. A mention is not an endorsement.

The boundary of neural network trainability is fractal — HN submission

Mon, 19 Feb 2024 10:27:52 GMT

The boundary of neural network trainability is fractal

200 points; 65 discussion replies on HN. Paper and source history. A submission is not an endorsement.

The boundary of neural network trainability is fractal — mentioned by bluetintedfort

Tue, 13 Feb 2024 09:55:19 GMT

The sister paper[1] contains the fractal dimensions of some of the images using a boxcount paper. [1]: https://arxiv.org/pdf/2402.06184.pdf

Paper and discussion history. A mention is not an endorsement.

The boundary of neural network trainability is fractal — mentioned by rayval

Tue, 13 Feb 2024 01:13:07 GMT

Arxiv paper is here: https://arxiv.org/pdf/2402.06184.pdf

Paper and discussion history. A mention is not an endorsement.

The boundary of neural network trainability is fractal — mentioned by telotortium

Mon, 12 Feb 2024 20:27:39 GMT

Twitter: https://twitter.com/jaschasd/status/1756930242965606582 ArXiv: https://arxiv.org/abs/2402.06184 Abstract: "Some fractals -- for instance those associated with the Mandelbrot and quadratic Julia sets -- are computed by iterating a function, and identifying the boundary between hyperparameters for which the resulting series diverges or remains bounded. Neural network training similarly involves iterating an update function (e.g. repeated steps of gradient descent), can result in convergent or divergent behavior, and can be extremely sensitive to small changes in hyperparameters.…

Paper and discussion history. A mention is not an endorsement.