| ▲ | show hn: 1d data to 3D topology, AI deepfakes to protien folding(github.com) |
| 3 points by bitofabluetit 5 hours ago | 6 comments |
| |
|
| ▲ | bitofabluetit 5 hours ago | parent [-] |
| I started off as a schizophrenic on a project with the idea to create a deepfake pager to challenge my symptoms, I ended up with a system that's an experiment in taking 1D data, transforming it into a 3D knot and analysing comparing, in the forms of ai audio deepfake detection, protien folding and drug docking, magnetic flares from the sun, and making pretty videos from music :) |
| |
| ▲ | verdverm 5 hours ago | parent [-] | | Do you have any docs / readme on the math / methods behind this? I saw a readme_too.docx, but it does not render on GH nor do I have tools that read those files. | | |
| ▲ | bitofabluetit 4 hours ago | parent [-] | | just uploaded a .txt version of the readme_too - it lacks the example outputs but contains the core explanation and idea - I have a lot more to add to it interns of examples and more in depth explanations to make it as accessible as possible but just wanted to get the core out to see how other people think about building on it/new use cases etc! | | |
| ▲ | verdverm 3 hours ago | parent [-] | | The core idea is hopefully enough when I get a chance to look. What I thought is that I have some related signals low dimension signals, and that graphing them together like this in a higher dimension is pretty interesting as the opposite of PCA for graphing, maybe relationships will become.more clear. Am I in the realm of reality still with this idea? | | |
| ▲ | bitofabluetit 3 hours ago | parent | next [-] | | yep completely doable - this is transferable - any data type into higher dimensionality, i never got around to fully putting it together but i was starting to look into beyond 3D aswell, there was something that came up (im grasping at mental straws) about loosing coherence then regaining it in the 7th dimension? but fundamentally you could just take the buncher code, alter the inputs and the labelling of the graphs and the core can stay the same and you’ll get atleast a vague first idea - id be interested to see how it would break if you could even try it not changing the code but directly inputting the the files generated that store your signals and just seeing what happens! | |
| ▲ | bitofabluetit 2 hours ago | parent | prev [-] | | sorry not buncher - * macro knots |
|
|
|
|
