undefined | Better HN

0 pointsxadhominemx1y ago0 comments

It’s not because of analytical convenience, it’s because of the central limit theorem.

0 comments

Not everything is a linear combination of large number of (IID) samples, and thus not everything is gaussian distributed.

kgwgk1y ago

You’re implying that many things are though.

jampekka1y ago

Yes, and I was explicit about it in another comment in this post.

kgwgk1y ago

Ok, so we all agree that in most cases the reason to minimize square error is that it’s the appropriate thing to minimize - not that it has an analytical solution, convenience or tradition.

j / k navigate · click thread line to collapse

0 comments

jampekka1y ago

Not everything is a linear combination of large number of (IID) samples, and thus not everything is gaussian distributed.

kgwgk1y ago

You’re implying that many things are though.

jampekka1y ago

Yes, and I was explicit about it in another comment in this post.

kgwgk1y ago

Ok, so we all agree that in most cases the reason to minimize square error is that it’s the appropriate thing to minimize - not that it has an analytical solution, convenience or tradition.

j / k navigate · click thread line to collapse