On Sun, Apr 26, 2026 at 02:11:43PM -0400, funkymail via cypherpunks wrote:
On Sun, Apr 26, 2026 at 02:09:37PM -0400, funkymail via cypherpunks wrote:
> > https://ar.anyone.tech/6cb-OyzPFi1nJYZ-uHAHH2uwoCbosf0AQuS_qGEgBXY#-U_3AcjRg...
https://ar.anyone.tech/yYCz9PNR58vqF0oJI0bivwVmiOJ4Nn3Bcelast6ynVw#j1dSORCMP... https://ar.anyone.tech/t_0UFmo0nYrMZrvyIqby3eTV-xSE4vgThsSqbi5FKjs
some upsetness the challenge is not for economic gain. so let's skip to this: you're not going to like it for more reasons than just that. but without all the influence it's just a simple algorithm task, it is a little deep inside weird intermediate-level areas of math and computer science and such. now i dunno why you might not like that stuff. but sometimes that can seem to be the issue.
and when we get to a spot, it's generally extremely hard for me and i start trying to push through it in spammy laborious microsteps. also normal for me, just a task closer to my inhibitions, likely cause i hold it as the litmus test.
https://ar.anyone.tech/Ktgtg2uyYl7bQqkEhApVkCGVZi0EpPRPMKU4QXEH4Oo#STcOlX11A... https://ar.anyone.tech/8qGktaPdZnrs1L4dgXW2gcHu6iQL-A6D_zRkBfWVCQc
ok wave 3 of starting.
In the paper they used FastICA, which is an approach to component analysis that uses maybe 3 existing matrix math concepts together in a feedback loop to construct mixing and unmixing matrices between measurements and signals. The RGB channels of skin color recombined to trend quickly and reliably to the user's puls
o8chies ouch
https://ar.anyone.tech/yz5qwtbaNxyukTrFyaO3YDFEcwweUyBvumiEDtvVnQg#x36BlkgBI... https://ar.anyone.tech/hcIYItxjB_waJHeNLQWXh-RDicVWKpMwanivZfucLAI
gotta a big strong stopping associated with "trend quickly and reliably to" might relate to stopping fast AI very strong
https://ar.anyone.tech/4JsidwtRdeBy8DCz0zj1jOPzjolbprJBTpcoGjLe7CQ#x36BlkgBI... https://ar.anyone.tech/ikIUTWDREYqAFVVMqz0Y62nGVft2_H1My2cC8sboLJE
but i got more optimization and tech techniques from the stopping :) new stuff: if you want to find a signal with a known kind of mixing, you can use _all_ the metrics you have that identify the signal (for FastICA the gaussianness is minimized to desum the signals, summed signals are more gaussian) and mathematically look at how the numbers of that metric directly relate to the mixing parameters (the derivative for example, somewhat corrupt lost part) to "invert" it. hard to say straight but the positivity is helpfup for trying to continue, got a really big stop though :s https://ar.anyone.tech/bEFskvOegFTcuLAV06d2baz1JE_BBNoInj2S5-_oncI#x36BlkgBI... https://ar.anyone.tech/N2xOI9J7Fsl_ZmPapL4WL6C4XHiKA7S1Ed6_DeXaC_c I am not affiliated with https://ar.anyone.tech .