Critiques

Very good work! Here are my scores, with comments below. I'm happy to discuss any points as needed.

Scale of 1 to 5, where 1 = strongly disagree and 5 = strongly agree:

---

• (5) The topic is relevant for the course.
• (4) The writing is clear and the English is good.
• (5) The page is written at an appropriate level for CPSC 522 students (where the students have diverse backgrounds).
• (4.5) The formalism (definitions, mathematics) was well chosen to make the page easier to understand.
• (4) The abstract is a concise and clear summary.
• (5) There were appropriate (original) examples that helped make the topic clear.
• (n/a) There was appropriate use of (pseudo-) code.
• (5) It had a good coverage of representations, semantics, inference and learning (as appropriate for the topic).
• (4.5) It is correct.
• (5) It was neither too short nor too long for the topic.
• (5) It was an appropriate unit for a page (it shouldn't be split into different topics or merged with another page).
• (4.5) It links to appropriate other pages in the wiki.
• (3) The references and links to external pages are well chosen.
• (5) I would recommend this page to someone who wanted to find out about the topic.
• (4.5) This page should be highlighted as an exemplary page for others to emulate.

Comments:

---

• A proofreading pass for grammar issues is recommended.
• One concern with using suicide as an example might be how students would react to it if they had been affected by suicide (if someone close to them committed suicide, or if they themselves had attempted suicide). I don't think the example should necessarily be changed; but I would recommend you make sure the tone of the writing isn't dismissive of the seriousness of the topic.
• It's good that you highlight significant terms using bold or italics; but that seems to happen more in the middle of the page than at the beginning or end. Another quick pass for significant terms to highlight might be beneficial.
• Some of your external links actually link to the 522 wiki, and so there is a way to format them as internal links (not having the arrow icon beside them). See the Help:Formatting page for more info. Also, if you want to link to a specific part of a 522 Wiki page (eg. Chain Rule), you can (and should) do that too.

Abstract:

• Feels more like an introduction than an abstract. Having it be somewhat shorter might be good.
• "Builds on" and "Related Pages": the instructions are still in the page (in italics); would probably be good to remove them.
• Related Pages: I'm not sure I agree with your description of Markov networks. My understanding is that they're in a separate category of graphical model that can represent some dependency structures that cannot be represented by BN (and vice versa).

Case Study:

• Using a single substantial example throughout to motivate and work through the material is a very interesting approach, and it seems to work well.

Definition:

• Again, I'm not sure the Markov network reference is correct (and it seems to contradict the description in Related Pages). I think it would be simpler to reference BN as a type of graphical model and link to that page instead.
• The set of random variables is used before it's "defined". I'd suggest moving the "definition" up a bit and applying subscripts for consistency with the rest of the section.

Dependencies in Bayesian Networks:

• I think the phrase "conditionally independent" should be used instead of simply "independent"; also, Common Descendant maybe should be Common Child, and "dependent" should be "conditionally dependent".

Constructing BN:

• The ToDo List approach was unexpected, but it is easy to follow and seems to work quite well. One recommendation is to ensure consistency in verb usage: 1. Define... 2. Apply... 3. (correct) 4. ...rewrite... 5. (correct)

Example:

• Long lists can be hard to read. Spaces after commas will help with this.
• If you can, you may want to make only the first letters of your variables capitalized so that long sequences of variables look less homogeneous.
• Applying Chain Rule: That looks like one big long equation, but the spacing between terms is inconsistent and the purpose behind the grouping of terms on each line is unclear. Same applies to Rewriting.
• Constructing the network: you have X_i - T, should be =; also, might want an external link for Machine Learning.
• Since you've used this one example throughout the page, it would be appropriate to put in some sample numbers and work out the sample solution: what is the probability of suicide in this Vancouver Police Dept case?

Hi Yaashaar, Thanks for the write up. It is indeed a pleasant read, given the way you have taken the bottom up approach. By motivating from an interesting real world example and then connecting that to the theory, which I found very engaging. However I have few suggestions that may further help in bettering it:

• Can you add a section on large scale Bayesian inference. I guess this is a very valid topic given current influence of big data.
• Adding some more references in bibliography, would be helpful to facilitate further reading. For e.g. you may give a little introduction to large scale bayesian and add pointers for detailed reading on that.
• A better formatting would help. For e.g. lines like the following should be formatted better:

In our example we would have required to specify $2^{1}2$ numbers for presenting the full joint probability distributions in the worst case scenario (fully connected network),however, by exploiting the conditional independences among variables through the Bayesian Network, we only defined:

$N\times D^{K}$

• Some of the sentences are very long and hard to follow. Better if the can be split to smaller ones. For e.g:

The idea behind the Bayesian Network and its power is where, by exploiting the conditional independency of a variable giving a subset of its predecessor nodes upon which the random variable directly depends, we can query as if we have the access over the full Joint Probability Distributions (JPD), while in fact, only conditional probability distributions were defined.

• A few quibbles apart from those that already been highlighted.

Bayesian Networks (also referred to as Belief Networks or Probabilistic Networks) is a --> should be are These case are from Death --> These cases

that directly affects variable --> affect

best, Prithu Banerjee

Hi Prithu,

Thank you very much for your careful inspection. I have addressed almost all of the issues you pointed out and I incorporated most of your kind suggestions. Just regarding the Bayesian inference in large scale, I believe that it should be classified under model Inferences, which were discussed in the Graphical Models page. In my opinion, a dedicated page for model inferences is required. With respect to the current size of the page, I would consider any discussion over the Bayesian Inference more than what has been out of the scope.

I really appreciate your comments.
Yaashaar

Hi jordon,

Thank you very much, your review helped me a lot. I have addressed the problems you brought to my attention. Also, as you suggested, I have added a section, to actually use the designed Bayesian Network, to find out whether or not the student committed suicide through inference by enumeration. Regarding the subject of the case study,I believe it shed a light on a very important matter among students. I've incorporated your suggestion where it shows how investigator managed to elicit the required information from the student's diary. I tried to keep the sarcasm and casual talks at its minimum in that area, while I was intentionally avoiding a formal-text-book like tone.

I really appreciate your detailed and section by section review.
Yaashaar