1.6 KiB
1.6 KiB
isHidden | menupriority | kind | created_at | title | author_name | author_uri | tags | ||||
---|---|---|---|---|---|---|---|---|---|---|---|
false | 1 | article | 2010-07-09T10:04:31+02:00 | Undecidabilities | Yann Esposito | yannesposito.com |
|
begindiv(intro)
<%= tldr %> Four different undecidabilities:
- Undecidability due to measure errors
- Hope: small measure error implies small predictive errors.
- Undecidability with big error from small measure error
- Hope: error remain controllable given a final mesure error
- Undecidability with discontinuities
- Hope: Undecidability is due to error, but without any measure error, all is decidable
- Undecidability without any measure error
Is the situation so bad we cannot truly believe anything? A theorem in learning theory (in my thesis) prove that under some reasonable conditions, we could learn many true things. But at no moment we can be absolutely sure this is the truth. But as far as I can see, we cannot do better than that.
enddiv
newcorps
The Undecidabilities
begindiv(intro)
enddiv
<%= leftblogimage("genese.png") %>
Undecidability due to measure errors
<%= leftblogimage('triangles_plats.png') %>
Growing errors Undecidability
<%= leftblogimage('trois_corps.png') %>
<%= leftblogimage('fractale.png') %>
<%= leftblogimage("Turing_undecidability.png") %>
Des bateaux à la rescousse de la vérité.
<%= leftblogimage("3DTriangle.png") %>
Fractions rationnelles
<%= leftblogimage("rational_fraction.png") %>