search

mml-book / mml-book.github.io

Companion webpage to the book "Mathematics For Machine Learning"
13.1k stars 2.42k forks source link

Typo in problem 5.9 #485

Closed carlolimb closed 4 years ago

carlolimb commented 4 years ago

5.9 We define g(z,v):= log p(x,z) - log q(z,v) z := t(e,v)

What is x in p(x,z)? It would make sense if it was a v and we considered p(v,z).

mpd37 commented 4 years ago

The definition is OK. The x does not play an important role in the context of the question. But we didn't want to reformulate it as p(z) alone as it resembles a problem in variational inference. Agree it's probably misleading, but I don't have a good workaround.

I took the example from here: https://media.nips.cc/Conferences/2016/Slides/6199-Slides.pdf

slide 75

While variational inference is far beyond the scope of the book, solving these kind of problems we can actually do.

carlolimb commented 4 years ago

Thanks for the quick response. Would it be prudent to state x \in \mathbb{R}^n and/or the dimension of the other variables?

On Tue, Feb 25, 2020 at 1:35 PM Marc Deisenroth notifications@github.com wrote:

The definition is OK. The x does not play an important role in the context of the question. But we didn't want to reformulate it as p(z) alone as it resembles a problem in variational inference. Agree it's probably misleading, but I don't have a good workaround.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/mml-book/mml-book.github.io/issues/485?email_source=notifications&email_token=AC6X3BY7OVLAHVP4ULOL2MTREVXIRA5CNFSM4K3QRYXKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEM5GLUI#issuecomment-591029713, or unsubscribe https://github.com/notifications/unsubscribe-auth/AC6X3B5MDQRSKZ2R7XWHILDREVXIRANCNFSM4K3QRYXA .

mpd37 commented 4 years ago

Good suggestion. I'll add this.