Comments (3)
First, please note that Figure 10.2 compares variational Bayes vs. expectation propagation (moment matching) or minimizing KL(q||p) vs. KL(p||q), and that the solutions have different properties as shown in the figure (see also Figure 10.3).
For either case, we do not need any assumption other than that q(Z) factorizes as (10.5) to obtain the spherical distributions as described in the text because the true distribution p(Z) has an elliptical distribution rotated 45 degrees so that the diagonal elements Lambda_11 and Lambda_22 of the precision matrix (and thus those of the covariance matrix) are equal.
I would suggest you run both the algorithms for some specific example by yourself to understand why.
from prml_errata.
I can understand your point and you are right. Thank you for your good comments. :-)
from prml_errata.
Hi Yousuke
I appreciate your good comments and excellent errata.
For this problem, I want to check one more thing with you.
As you commented, I can understand \Lambda_{11} = \Lambda_{22} and the solution has
spherical Gaussian.
However, I think that as author showed, both solutions based on (10.9) and KL(p||q) are same as following.
https://drive.google.com/file/d/1OKM4T6SL1-3zPfqnXcy3TSEXbrup4T6r/view?usp=sharing
So, I think that only one figure in Figure 10.2 is relevant to this solution and cannot have
two different cases.
Is my understanding correct?
Thank you.
from prml_errata.
Related Issues (18)
- Dismissal of "bayesian methods don't overfit" is too quick HOT 11
- Missing x in equation 3.5 HOT 1
- Link to official support page broken
- Sylvster's determinant identity (C.14) is not a consequence of the push-through identity (C.6); use the Schur complement instead
- Error in eq. (3.99) HOT 2
- Self-contained proof for Wishart distribution
- I believe M should be 4 in 5.5. Regularization in Neural Networks, p256
- Errata for positive definite matrix in page 701 HOT 2
- Errata fixed by Bishop is correct at page 101? HOT 2
- Errata at page 35 in More PRML Errata
- Introduce paragraphs and paragraph headers
- Add some more properties of log minus digamma function
- Expand on robust regression
- Give some motivation for Fisher information matrix
- Add errata for some hyphenation inconsistencies HOT 1
- Kronecker's delta used w/o definition
- Success probability $\mu$ for Bernoulli should be restricted on $(0, 1)$
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from prml_errata.